Datasets:

librarian-bot

/

dataset_abstracts

like 1

{ "abstract": "We introduce WikiLingua, a large-scale, multilingual dataset for the\nevaluation of crosslingual abstractive summarization systems. We extract\narticle and summary pairs in 18 languages from WikiHow, a high quality,\ncollaborative resource of how-to guides on a diverse set of topics written by\nhuman authors. We create gold-standard article-summary alignments across\nlanguages by aligning the images that are used to describe each how-to step in\nan article. As a set of baselines for further studies, we evaluate the\nperformance of existing cross-lingual abstractive summarization methods on our\ndataset. We further propose a method for direct crosslingual summarization\n(i.e., without requiring translation at inference time) by leveraging synthetic\ndata and Neural Machine Translation as a pre-training step. Our method\nsignificantly outperforms the baseline approaches, while being more cost\nefficient during inference.", "title": "WikiLingua: A New Benchmark Dataset for Cross-Lingual Abstractive Summarization", "url": "http://arxiv.org/abs/2010.03093v1" }

{ "text_length": 1034 }

{ "abstract": "Purpose: Hard-to-interpret Black-box Machine Learning (ML) were often used\nfor early Alzheimer's Disease (AD) detection.\n Methods: To interpret eXtreme Gradient Boosting (XGBoost), Random Forest\n(RF), and Support Vector Machine (SVM) black-box models a workflow based on\nShapley values was developed. All models were trained on the Alzheimer's\nDisease Neuroimaging Initiative (ADNI) dataset and evaluated for an independent\nADNI test set, as well as the external Australian Imaging and Lifestyle\nflagship study of Ageing (AIBL), and Open Access Series of Imaging Studies\n(OASIS) datasets. Shapley values were compared to intuitively interpretable\nDecision Trees (DTs), and Logistic Regression (LR), as well as natural and\npermutation feature importances. To avoid the reduction of the explanation\nvalidity caused by correlated features, forward selection and aspect\nconsolidation were implemented.\n Results: Some black-box models outperformed DTs and LR. The forward-selected\nfeatures correspond to brain areas previously associated with AD. Shapley\nvalues identified biologically plausible associations with moderate to strong\ncorrelations with feature importances. The most important RF features to\npredict AD conversion were the volume of the amygdalae, and a cognitive test\nscore. Good cognitive test performances and large brain volumes decreased the\nAD risk. The models trained using cognitive test scores significantly\noutperformed brain volumetric models ($p<0.05$). Cognitive Normal (CN) vs. AD\nmodels were successfully transferred to external datasets.\n Conclusion: In comparison to previous work, improved performances for ADNI\nand AIBL were achieved for CN vs. Mild Cognitive Impairment (MCI)\nclassification using brain volumes. The Shapley values and the feature\nimportances showed moderate to strong correlations.", "title": "Machine Learning Workflow to Explain Black-box Models for Early Alzheimer's Disease Classification Evaluated for Multiple Datasets", "url": "http://arxiv.org/abs/2205.05907v2" }

{ "text_length": 1995 }

{ "abstract": "This study presents machine learning models that forecast and categorize lost\ncirculation severity preemptively using a large class imbalanced drilling\ndataset. We demonstrate reproducible core techniques involved in tackling a\nlarge drilling engineering challenge utilizing easily interpretable machine\nlearning approaches.\n We utilized a 65,000+ records data with class imbalance problem from Azadegan\noilfield formations in Iran. Eleven of the dataset's seventeen parameters are\nchosen to be used in the classification of five lost circulation events. To\ngenerate classification models, we used six basic machine learning algorithms\nand four ensemble learning methods. Linear Discriminant Analysis (LDA),\nLogistic Regression (LR), Support Vector Machines (SVM), Classification and\nRegression Trees (CART), k-Nearest Neighbors (KNN), and Gaussian Naive Bayes\n(GNB) are the six fundamental techniques. We also used bagging and boosting\nensemble learning techniques in the investigation of solutions for improved\npredicting performance. The performance of these algorithms is measured using\nfour metrics: accuracy, precision, recall, and F1-score. The F1-score weighted\nto represent the data imbalance is chosen as the preferred evaluation\ncriterion.\n The CART model was found to be the best in class for identifying drilling\nfluid circulation loss events with an average weighted F1-score of 0.9904 and\nstandard deviation of 0.0015. Upon application of ensemble learning techniques,\na Random Forest ensemble of decision trees showed the best predictive\nperformance. It identified and classified lost circulation events with a\nperfect weighted F1-score of 1.0. Using Permutation Feature Importance (PFI),\nthe measured depth was found to be the most influential factor in accurately\nrecognizing lost circulation events while drilling.", "title": "A Case Study on the Classification of Lost Circulation Events During Drilling using Machine Learning Techniques on an Imbalanced Large Dataset", "url": "http://arxiv.org/abs/2209.01607v2" }

{ "text_length": 2012 }

{ "abstract": "Movie-making has become one of the most costly and risky endeavors in the\nentertainment industry. Continuous change in the preference of the audience\nmakes it harder to predict what kind of movie will be financially successful at\nthe box office. So, it is no wonder that cautious, intelligent stakeholders and\nlarge production houses will always want to know the probable revenue that will\nbe generated by a movie before making an investment. Researchers have been\nworking on finding an optimal strategy to help investors in making the right\ndecisions. But the lack of a large, up-to-date dataset makes their work harder.\nIn this work, we introduce an up-to-date, richer, and larger dataset that we\nhave prepared by scraping IMDb for researchers and data analysts to work with.\nThe compiled dataset contains the summery data of 7.5 million titles and detail\ninformation of more than 200K movies. Additionally, we perform different\nstatistical analysis approaches on our dataset to find out how a movie's\nrevenue is affected by different pre-released attributes such as budget,\nruntime, release month, content rating, genre etc. In our analysis, we have\nfound that having a star cast/director has a positive impact on generated\nrevenue. We introduce a novel approach for calculating the star power of a\nmovie. Based on our analysis we select a set of attributes as features and\ntrain different machine learning algorithms to predict a movie's expected\nrevenue. Based on generated revenue, we classified the movies in 10 categories\nand achieved a one-class-away accuracy rate of almost 60% (bingo accuracy of\n30%). All the generated datasets and analysis codes are available online. We\nalso made the source codes of our scraper bots public, so that researchers\ninterested in extending this work can easily modify these bots as they need and\nprepare their own up-to-date datasets.", "title": "Presenting a Larger Up-to-date Movie Dataset and Investigating the Effects of Pre-released Attributes on Gross Revenue", "url": "http://arxiv.org/abs/2110.07039v2" }

{ "text_length": 2030 }

{ "abstract": "Fighting online hate speech is a challenge that is usually addressed using\nNatural Language Processing via automatic detection and removal of hate\ncontent. Besides this approach, counter narratives have emerged as an effective\ntool employed by NGOs to respond to online hate on social media platforms. For\nthis reason, Natural Language Generation is currently being studied as a way to\nautomatize counter narrative writing. However, the existing resources necessary\nto train NLG models are limited to 2-turn interactions (a hate speech and a\ncounter narrative as response), while in real life, interactions can consist of\nmultiple turns. In this paper, we present a hybrid approach for dialogical data\ncollection, which combines the intervention of human expert annotators over\nmachine generated dialogues obtained using 19 different configurations. The\nresult of this work is DIALOCONAN, the first dataset comprising over 3000\nfictitious multi-turn dialogues between a hater and an NGO operator, covering 6\ntargets of hate.", "title": "Human-Machine Collaboration Approaches to Build a Dialogue Dataset for Hate Speech Countering", "url": "http://arxiv.org/abs/2211.03433v1" }

{ "text_length": 1152 }

{ "abstract": "We present MultiCoNER, a large multilingual dataset for Named Entity\nRecognition that covers 3 domains (Wiki sentences, questions, and search\nqueries) across 11 languages, as well as multilingual and code-mixing subsets.\nThis dataset is designed to represent contemporary challenges in NER, including\nlow-context scenarios (short and uncased text), syntactically complex entities\nlike movie titles, and long-tail entity distributions. The 26M token dataset is\ncompiled from public resources using techniques such as heuristic-based\nsentence sampling, template extraction and slotting, and machine translation.\nWe applied two NER models on our dataset: a baseline XLM-RoBERTa model, and a\nstate-of-the-art GEMNET model that leverages gazetteers. The baseline achieves\nmoderate performance (macro-F1=54%), highlighting the difficulty of our data.\nGEMNET, which uses gazetteers, improvement significantly (average improvement\nof macro-F1=+30%). MultiCoNER poses challenges even for large pre-trained\nlanguage models, and we believe that it can help further research in building\nrobust NER systems. MultiCoNER is publicly available at\nhttps://registry.opendata.aws/multiconer/ and we hope that this resource will\nhelp advance research in various aspects of NER.", "title": "MultiCoNER: A Large-scale Multilingual dataset for Complex Named Entity Recognition", "url": "http://arxiv.org/abs/2208.14536v1" }

{ "text_length": 1375 }

{ "abstract": "Machine learning (ML) approaches have demonstrated promising results in a\nwide range of healthcare applications. Data plays a crucial role in developing\nML-based healthcare systems that directly affect people's lives. Many of the\nethical issues surrounding the use of ML in healthcare stem from structural\ninequalities underlying the way we collect, use, and handle data. Developing\nguidelines to improve documentation practices regarding the creation, use, and\nmaintenance of ML healthcare datasets is therefore of critical importance. In\nthis work, we introduce Healthsheet, a contextualized adaptation of the\noriginal datasheet questionnaire ~\\cite{gebru2018datasheets} for\nhealth-specific applications. Through a series of semi-structured interviews,\nwe adapt the datasheets for healthcare data documentation. As part of the\nHealthsheet development process and to understand the obstacles researchers\nface in creating datasheets, we worked with three publicly-available healthcare\ndatasets as our case studies, each with different types of structured data:\nElectronic health Records (EHR), clinical trial study data, and\nsmartphone-based performance outcome measures. Our findings from the\ninterviewee study and case studies show 1) that datasheets should be\ncontextualized for healthcare, 2) that despite incentives to adopt\naccountability practices such as datasheets, there is a lack of consistency in\nthe broader use of these practices 3) how the ML for health community views\ndatasheets and particularly \\textit{Healthsheets} as diagnostic tool to surface\nthe limitations and strength of datasets and 4) the relative importance of\ndifferent fields in the datasheet to healthcare concerns.", "title": "Healthsheet: Development of a Transparency Artifact for Health Datasets", "url": "http://arxiv.org/abs/2202.13028v1" }

{ "text_length": 1803 }

{ "abstract": "The need for secure Internet of Things (IoT) devices is growing as IoT\ndevices are becoming more integrated into vital networks. Many systems rely on\nthese devices to remain available and provide reliable service. Denial of\nservice attacks against IoT devices are a real threat due to the fact these low\npower devices are very susceptible to denial-of-service attacks. Machine\nlearning enabled network intrusion detection systems are effective at\nidentifying new threats, but they require a large amount of data to work well.\nThere are many network traffic data sets but very few that focus on IoT network\ntraffic. Within the IoT network data sets there is a lack of CoAP denial of\nservice data. We propose a novel data set covering this gap. We develop a new\ndata set by collecting network traffic from real CoAP denial of service attacks\nand compare the data on multiple different machine learning classifiers. We\nshow that the data set is effective on many classifiers.", "title": "CoAP-DoS: An IoT Network Intrusion Dataset", "url": "http://arxiv.org/abs/2206.14341v1" }

{ "text_length": 1049 }

{ "abstract": "This paper describes a multi-feature dataset for training machine learning\nclassifiers for detecting malicious Windows Portable Executable (PE) files. The\ndataset includes four feature sets from 18,551 binary samples belonging to five\nmalware families including Spyware, Ransomware, Downloader, Backdoor and\nGeneric Malware. The feature sets include the list of DLLs and their functions,\nvalues of different fields of PE Header and Sections. First, we explain the\ndata collection and creation phase and then we explain how did we label the\nsamples in it using VirusTotal's services. Finally, we explore the dataset to\ndescribe how this dataset can benefit the researchers for static malware\nanalysis. The dataset is made public in the hope that it will help inspire\nmachine learning research for malware detection.", "title": "Multi-feature Dataset for Windows PE Malware Classification", "url": "http://arxiv.org/abs/2210.16285v1" }

{ "text_length": 908 }

{ "abstract": "There is a limited amount of publicly available data to support research in\nmalware analysis technology. Particularly, there are virtually no publicly\navailable datasets generated from rich sandboxes such as Cuckoo/CAPE. The\nbenefit of using dynamic sandboxes is the realistic simulation of file\nexecution in the target machine and obtaining a log of such execution. The\nmachine can be infected by malware hence there is a good chance of capturing\nthe malicious behavior in the execution logs, thus allowing researchers to\nstudy such behavior in detail. Although the subsequent analysis of log\ninformation is extensively covered in industrial cybersecurity backends, to our\nknowledge there has been only limited effort invested in academia to advance\nsuch log analysis capabilities using cutting edge techniques. We make this\nsample dataset available to support designing new machine learning methods for\nmalware detection, especially for automatic detection of generic malicious\nbehavior. The dataset has been collected in cooperation between Avast Software\nand Czech Technical University - AI Center (AIC).", "title": "Avast-CTU Public CAPE Dataset", "url": "http://arxiv.org/abs/2209.03188v1" }

{ "text_length": 1172 }

{ "abstract": "The data article describes the Road Damage Dataset, RDD2022, which comprises\n47,420 road images from six countries, Japan, India, the Czech Republic,\nNorway, the United States, and China. The images have been annotated with more\nthan 55,000 instances of road damage. Four types of road damage, namely\nlongitudinal cracks, transverse cracks, alligator cracks, and potholes, are\ncaptured in the dataset. The annotated dataset is envisioned for developing\ndeep learning-based methods to detect and classify road damage automatically.\nThe dataset has been released as a part of the Crowd sensing-based Road Damage\nDetection Challenge (CRDDC2022). The challenge CRDDC2022 invites researchers\nfrom across the globe to propose solutions for automatic road damage detection\nin multiple countries. The municipalities and road agencies may utilize the\nRDD2022 dataset, and the models trained using RDD2022 for low-cost automatic\nmonitoring of road conditions. Further, computer vision and machine learning\nresearchers may use the dataset to benchmark the performance of different\nalgorithms for other image-based applications of the same type (classification,\nobject detection, etc.).", "title": "RDD2022: A multi-national image dataset for automatic Road Damage Detection", "url": "http://arxiv.org/abs/2209.08538v1" }

{ "text_length": 1284 }

{ "abstract": "Designing robust machine learning systems remains an open problem, and there\nis a need for benchmark problems that cover both environmental changes and\nevaluation on a downstream task. In this work, we introduce AVOIDDS, a\nrealistic object detection benchmark for the vision-based aircraft\ndetect-and-avoid problem. We provide a labeled dataset consisting of 72,000\nphotorealistic images of intruder aircraft with various lighting conditions,\nweather conditions, relative geometries, and geographic locations. We also\nprovide an interface that evaluates trained models on slices of this dataset to\nidentify changes in performance with respect to changing environmental\nconditions. Finally, we implement a fully-integrated, closed-loop simulator of\nthe vision-based detect-and-avoid problem to evaluate trained models with\nrespect to the downstream collision avoidance task. This benchmark will enable\nfurther research in the design of robust machine learning systems for use in\nsafety-critical applications. The AVOIDDS dataset and code are publicly\navailable at\n$\\href{https://purl.stanford.edu/hj293cv5980}{purl.stanford.edu/hj293cv5980}$\nand\n$\\href{https://github.com/sisl/VisionBasedAircraftDAA}{github.com/sisl/VisionBasedAircraftDAA}$,\nrespectively.", "title": "AVOIDDS: Aircraft Vision-based Intruder Detection Dataset and Simulator", "url": "http://arxiv.org/abs/2306.11203v1" }

{ "text_length": 1361 }

{ "abstract": "Purpose: We propose a formal framework for the modeling and segmentation of\nminimally-invasive surgical tasks using a unified set of motion primitives\n(MPs) to enable more objective labeling and the aggregation of different\ndatasets.\n Methods: We model dry-lab surgical tasks as finite state machines,\nrepresenting how the execution of MPs as the basic surgical actions results in\nthe change of surgical context, which characterizes the physical interactions\namong tools and objects in the surgical environment. We develop methods for\nlabeling surgical context based on video data and for automatic translation of\ncontext to MP labels. We then use our framework to create the COntext and\nMotion Primitive Aggregate Surgical Set (COMPASS), including six dry-lab\nsurgical tasks from three publicly-available datasets (JIGSAWS, DESK, and\nROSMA), with kinematic and video data and context and MP labels.\n Results: Our context labeling method achieves near-perfect agreement between\nconsensus labels from crowd-sourcing and expert surgeons. Segmentation of tasks\nto MPs results in the creation of the COMPASS dataset that nearly triples the\namount of data for modeling and analysis and enables the generation of separate\ntranscripts for the left and right tools.\n Conclusion: The proposed framework results in high quality labeling of\nsurgical data based on context and fine-grained MPs. Modeling surgical tasks\nwith MPs enables the aggregation of different datasets and the separate\nanalysis of left and right hands for bimanual coordination assessment. Our\nformal framework and aggregate dataset can support the development of\nexplainable and multi-granularity models for improved surgical process\nanalysis, skill assessment, error detection, and autonomy.", "title": "COMPASS: A Formal Framework and Aggregate Dataset for Generalized Surgical Procedure Modeling", "url": "http://arxiv.org/abs/2209.06424v5" }

{ "text_length": 1884 }

{ "abstract": "Previous work on learning physical systems from data has focused on\nhigh-resolution grid-structured measurements. However, real-world knowledge of\nsuch systems (e.g. weather data) relies on sparsely scattered measuring\nstations. In this paper, we introduce a novel simulated benchmark dataset,\nDynaBench, for learning dynamical systems directly from sparsely scattered data\nwithout prior knowledge of the equations. The dataset focuses on predicting the\nevolution of a dynamical system from low-resolution, unstructured measurements.\nWe simulate six different partial differential equations covering a variety of\nphysical systems commonly used in the literature and evaluate several machine\nlearning models, including traditional graph neural networks and point cloud\nprocessing models, with the task of predicting the evolution of the system. The\nproposed benchmark dataset is expected to advance the state of art as an\nout-of-the-box easy-to-use tool for evaluating models in a setting where only\nunstructured low-resolution observations are available. The benchmark is\navailable at https://anonymous.4open.science/r/code-2022-dynabench/.", "title": "DynaBench: A benchmark dataset for learning dynamical systems from low-resolution data", "url": "http://arxiv.org/abs/2306.05805v2" }

{ "text_length": 1261 }

{ "abstract": "Accurate channel knowledge is critical in massive multiple-input\nmultiple-output (MIMO), which motivates the use of channel prediction. Machine\nlearning techniques for channel prediction hold much promise, but current\nschemes are limited in their ability to adapt to changes in the environment\nbecause they require large training overheads. To accurately predict wireless\nchannels for new environments with reduced training overhead, we propose a fast\nadaptive channel prediction technique based on a meta-learning algorithm for\nmassive MIMO communications. We exploit the model-agnostic meta-learning (MAML)\nalgorithm to achieve quick adaptation with a small amount of labeled data.\nAlso, to improve the prediction accuracy, we adopt the denoising process for\nthe training data by using deep image prior (DIP). Numerical results show that\nthe proposed MAML-based channel predictor can improve the prediction accuracy\nwith only a few fine-tuning samples. The DIP-based denoising process gives an\nadditional gain in channel prediction, especially in low signal-to-noise ratio\nregimes.", "title": "Massive MIMO Channel Prediction Via Meta-Learning and Deep Denoising: Is a Small Dataset Enough?", "url": "http://arxiv.org/abs/2210.08770v1" }

{ "text_length": 1214 }

{ "abstract": "We introduce TTSWING, a novel dataset designed for table tennis swing\nanalysis. This dataset comprises comprehensive swing information obtained\nthrough 9-axis sensors integrated into custom-made racket grips, accompanied by\nanonymized demographic data of the players. We detail the data collection and\nannotation procedures. Furthermore, we conduct pilot studies utilizing diverse\nmachine learning models for swing analysis. TTSWING holds tremendous potential\nto facilitate innovative research in table tennis analysis and is a valuable\nresource for the scientific community. We release the dataset and experimental\ncodes at https://github.com/DEPhantom/TTSWING.", "title": "TTSWING: a Dataset for Table Tennis Swing Analysis", "url": "http://arxiv.org/abs/2306.17550v1" }

{ "text_length": 747 }

{ "abstract": "Rapid development in deep learning model construction has prompted an\nincreased need for appropriate training data. The popularity of large datasets\n- sometimes known as \"big data\" - has diverted attention from assessing their\nquality. Training on large datasets often requires excessive system resources\nand an infeasible amount of time. Furthermore, the supervised machine learning\nprocess has yet to be fully automated: for supervised learning, large datasets\nrequire more time for manually labeling samples. We propose a method of\ncurating smaller datasets with comparable out-of-distribution model accuracy\nafter an initial training session using an appropriate distribution of samples\nclassified by how difficult it is for a model to learn from them.", "title": "Evaluating and Crafting Datasets Effective for Deep Learning With Data Maps", "url": "http://arxiv.org/abs/2208.10033v2" }

{ "text_length": 866 }

{ "abstract": "With increasingly more data and computation involved in their training,\nmachine learning models constitute valuable intellectual property. This has\nspurred interest in model stealing, which is made more practical by advances in\nlearning with partial, little, or no supervision. Existing defenses focus on\ninserting unique watermarks in a model's decision surface, but this is\ninsufficient: the watermarks are not sampled from the training distribution and\nthus are not always preserved during model stealing. In this paper, we make the\nkey observation that knowledge contained in the stolen model's training set is\nwhat is common to all stolen copies. The adversary's goal, irrespective of the\nattack employed, is always to extract this knowledge or its by-products. This\ngives the original model's owner a strong advantage over the adversary: model\nowners have access to the original training data. We thus introduce $dataset$\n$inference$, the process of identifying whether a suspected model copy has\nprivate knowledge from the original model's dataset, as a defense against model\nstealing. We develop an approach for dataset inference that combines\nstatistical testing with the ability to estimate the distance of multiple data\npoints to the decision boundary. Our experiments on CIFAR10, SVHN, CIFAR100 and\nImageNet show that model owners can claim with confidence greater than 99% that\ntheir model (or dataset as a matter of fact) was stolen, despite only exposing\n50 of the stolen model's training points. Dataset inference defends against\nstate-of-the-art attacks even when the adversary is adaptive. Unlike prior\nwork, it does not require retraining or overfitting the defended model.", "title": "Dataset Inference: Ownership Resolution in Machine Learning", "url": "http://arxiv.org/abs/2104.10706v1" }

{ "text_length": 1786 }

{ "abstract": "The paper describes the MetroPT data set, an outcome of a eXplainable\nPredictive Maintenance (XPM) project with an urban metro public transportation\nservice in Porto, Portugal. The data was collected in 2022 that aimed to\nevaluate machine learning methods for online anomaly detection and failure\nprediction. By capturing several analogic sensor signals (pressure,\ntemperature, current consumption), digital signals (control signals, discrete\nsignals), and GPS information (latitude, longitude, and speed), we provide a\ndataset that can be easily used to evaluate online machine learning methods.\nThis dataset contains some interesting characteristics and can be a good\nbenchmark for predictive maintenance models.", "title": "A Benchmark dataset for predictive maintenance", "url": "http://arxiv.org/abs/2207.05466v3" }

{ "text_length": 795 }

{ "abstract": "Datasets scraped from the internet have been critical to the successes of\nlarge-scale machine learning. Yet, this very success puts the utility of future\ninternet-derived datasets at potential risk, as model outputs begin to replace\nhuman annotations as a source of supervision.\n In this work, we first formalize a system where interactions with one model\nare recorded as history and scraped as training data in the future. We then\nanalyze its stability over time by tracking changes to a test-time bias\nstatistic (e.g. gender bias of model predictions). We find that the degree of\nbias amplification is closely linked to whether the model's outputs behave like\nsamples from the training distribution, a behavior which we characterize and\ndefine as consistent calibration. Experiments in three conditional prediction\nscenarios - image classification, visual role-labeling, and language generation\n- demonstrate that models that exhibit a sampling-like behavior are more\ncalibrated and thus more stable. Based on this insight, we propose an\nintervention to help calibrate and stabilize unstable feedback systems.\n Code is available at https://github.com/rtaori/data_feedback.", "title": "Data Feedback Loops: Model-driven Amplification of Dataset Biases", "url": "http://arxiv.org/abs/2209.03942v1" }

{ "text_length": 1276 }

{ "abstract": "Numerical simulations of Earth's weather and climate require substantial\namounts of computation. This has led to a growing interest in replacing\nsubroutines that explicitly compute physical processes with approximate machine\nlearning (ML) methods that are fast at inference time. Within weather and\nclimate models, atmospheric radiative transfer (RT) calculations are especially\nexpensive. This has made them a popular target for neural network-based\nemulators. However, prior work is hard to compare due to the lack of a\ncomprehensive dataset and standardized best practices for ML benchmarking. To\nfill this gap, we build a large dataset, ClimART, with more than \\emph{10\nmillion samples from present, pre-industrial, and future climate conditions},\nbased on the Canadian Earth System Model. ClimART poses several methodological\nchallenges for the ML community, such as multiple out-of-distribution test\nsets, underlying domain physics, and a trade-off between accuracy and inference\nspeed. We also present several novel baselines that indicate shortcomings of\ndatasets and network architectures used in prior work. Download instructions,\nbaselines, and code are available at: https://github.com/RolnickLab/climart", "title": "ClimART: A Benchmark Dataset for Emulating Atmospheric Radiative Transfer in Weather and Climate Models", "url": "http://arxiv.org/abs/2111.14671v1" }

{ "text_length": 1354 }

{ "abstract": "Purpose: Object detection is rapidly evolving through machine learning\ntechnology in automation systems. Well prepared data is necessary to train the\nalgorithms. Accordingly, the objective of this paper is to describe a\nre-evaluation of the so-called Logistics Objects in Context (LOCO) dataset,\nwhich is the first dataset for object detection in the field of intralogistics.\n Methodology: We use an experimental research approach with three steps to\nevaluate the LOCO dataset. Firstly, the images on GitHub were analyzed to\nunderstand the dataset better. Secondly, Google Drive Cloud was used for\ntraining purposes to revisit the algorithmic implementation and training.\nLastly, the LOCO dataset was examined, if it is possible to achieve the same\ntraining results in comparison to the original publications.\n Findings: The mean average precision, a common benchmark in object detection,\nachieved in our study was 64.54%, and shows a significant increase from the\ninitial study of the LOCO authors, achieving 41%. However, improvement\npotential is seen specifically within object types of forklifts and pallet\ntruck.\n Originality: This paper presents the first critical replication study of the\nLOCO dataset for object detection in intralogistics. It shows that the training\nwith better hyperparameters based on LOCO can even achieve a higher accuracy\nthan presented in the original publication. However, there is also further room\nfor improving the LOCO dataset.", "title": "Critical Evaluation of LOCO dataset with Machine Learning", "url": "http://arxiv.org/abs/2209.13499v1" }

{ "text_length": 1559 }

{ "abstract": "As our ability to sense increases, we are experiencing a transition from\ndata-poor problems, in which the central issue is a lack of relevant data, to\ndata-rich problems, in which the central issue is to identify a few relevant\nfeatures in a sea of observations. Motivated by applications in\ngravitational-wave astrophysics, we study the problem of predicting the\npresence of transient noise artifacts in a gravitational wave detector from a\nrich collection of measurements from the detector and its environment. We argue\nthat feature learning--in which relevant features are optimized from data--is\ncritical to achieving high accuracy. We introduce models that reduce the error\nrate by over 60% compared to the previous state of the art, which used fixed,\nhand-crafted features. Feature learning is useful not only because it improves\nperformance on prediction tasks; the results provide valuable information about\npatterns associated with phenomena of interest that would otherwise be\nundiscoverable. In our application, features found to be associated with\ntransient noise provide diagnostic information about its origin and suggest\nmitigation strategies. Learning in high-dimensional settings is challenging.\nThrough experiments with a variety of architectures, we identify two key\nfactors in successful models: sparsity, for selecting relevant variables within\nthe high-dimensional observations; and depth, which confers flexibility for\nhandling complex interactions and robustness with respect to temporal\nvariations. We illustrate their significance through systematic experiments on\nreal detector data. Our results provide experimental corroboration of common\nassumptions in the machine-learning community and have direct applicability to\nimproving our ability to sense gravitational waves, as well as to many other\nproblem settings with similarly high-dimensional, noisy, or partly irrelevant\ndata.", "title": "Architectural Optimization and Feature Learning for High-Dimensional Time Series Datasets", "url": "http://arxiv.org/abs/2202.13486v2" }

{ "text_length": 2031 }

{ "abstract": "Knowledge is central to human and scientific developments. Natural Language\nProcessing (NLP) allows automated analysis and creation of knowledge. Data is a\ncrucial NLP and machine learning ingredient. The scarcity of open datasets is a\nwell-known problem in machine and deep learning research. This is very much the\ncase for textual NLP datasets in English and other major world languages. For\nthe Bangla language, the situation is even more challenging and the number of\nlarge datasets for NLP research is practically nil. We hereby present Potrika,\na large single-label Bangla news article textual dataset curated for NLP\nresearch from six popular online news portals in Bangladesh (Jugantor,\nJaijaidin, Ittefaq, Kaler Kontho, Inqilab, and Somoyer Alo) for the period\n2014-2020. The articles are classified into eight distinct categories\n(National, Sports, International, Entertainment, Economy, Education, Politics,\nand Science \\& Technology) providing five attributes (News Article, Category,\nHeadline, Publication Date, and Newspaper Source). The raw dataset contains\n185.51 million words and 12.57 million sentences contained in 664,880 news\narticles. Moreover, using NLP augmentation techniques, we create from the raw\n(unbalanced) dataset another (balanced) dataset comprising 320,000 news\narticles with 40,000 articles in each of the eight news categories. Potrika\ncontains both the datasets (raw and balanced) to suit a wide range of NLP\nresearch. By far, to the best of our knowledge, Potrika is the largest and the\nmost extensive dataset for news classification.", "title": "Potrika: Raw and Balanced Newspaper Datasets in the Bangla Language with Eight Topics and Five Attributes", "url": "http://arxiv.org/abs/2210.09389v1" }

{ "text_length": 1714 }

{ "abstract": "The development of machine learning models for electrocatalysts requires a\nbroad set of training data to enable their use across a wide variety of\nmaterials. One class of materials that currently lacks sufficient training data\nis oxides, which are critical for the development of OER catalysts. To address\nthis, we developed the OC22 dataset, consisting of 62,331 DFT relaxations\n(~9,854,504 single point calculations) across a range of oxide materials,\ncoverages, and adsorbates. We define generalized total energy tasks that enable\nproperty prediction beyond adsorption energies; we test baseline performance of\nseveral graph neural networks; and we provide pre-defined dataset splits to\nestablish clear benchmarks for future efforts. In the most general task,\nGemNet-OC sees a ~36% improvement in energy predictions when combining the\nchemically dissimilar OC20 and OC22 datasets via fine-tuning. Similarly, we\nachieved a ~19% improvement in total energy predictions on OC20 and a ~9%\nimprovement in force predictions in OC22 when using joint training. We\ndemonstrate the practical utility of a top performing model by capturing\nliterature adsorption energies and important OER scaling relationships. We\nexpect OC22 to provide an important benchmark for models seeking to incorporate\nintricate long-range electrostatic and magnetic interactions in oxide surfaces.\nDataset and baseline models are open sourced, and a public leaderboard is\navailable to encourage continued community developments on the total energy\ntasks and data.", "title": "The Open Catalyst 2022 (OC22) Dataset and Challenges for Oxide Electrocatalysts", "url": "http://arxiv.org/abs/2206.08917v3" }

{ "text_length": 1646 }

{ "abstract": "In this work we present BrazilDAM, a novel public dataset based on Sentinel-2\nand Landsat-8 satellite images covering all tailings dams cataloged by the\nBrazilian National Mining Agency (ANM). The dataset was built using\ngeoreferenced images from 769 dams, recorded between 2016 and 2019. The time\nseries were processed in order to produce cloud free images. The dams contain\nmining waste from different ore categories and have highly varying shapes,\nareas and volumes, making BrazilDAM particularly interesting and challenging to\nbe used in machine learning benchmarks. The original catalog contains, besides\nthe dam coordinates, information about: the main ore, constructive method, risk\ncategory, and associated potential damage. To evaluate BrazilDAM's predictive\npotential we performed classification essays using state-of-the-art deep\nConvolutional Neural Network (CNNs). In the experiments, we achieved an average\nclassification accuracy of 94.11% in tailing dam binary classification task. In\naddition, others four setups of experiments were made using the complementary\ninformation from the original catalog, exhaustively exploiting the capacity of\nthe proposed dataset.", "title": "BrazilDAM: A Benchmark dataset for Tailings Dam Detection", "url": "http://arxiv.org/abs/2003.07948v2" }

{ "text_length": 1271 }

{ "abstract": "A generative adversarial network (GAN) is a class of machine learning\nframeworks designed by Goodfellow et al. in 2014. In the GAN framework, the\ngenerative model is pitted against an adversary: a discriminative model that\nlearns to determine whether a sample is from the model distribution or the data\ndistribution. GAN generates new samples from the same distribution as the\ntraining set. In this work, we aim to generate a new dataset that has a\ndifferent distribution from the training set. In addition, the Jensen-Shannon\ndivergence between the distributions of the generative and training datasets\ncan be controlled by some target $\\delta \\in [0, 1]$. Our work is motivated by\napplications in generating new kinds of rice that have similar characteristics\nas good rice.", "title": "Generative Adversarial Nets: Can we generate a new dataset based on only one training set?", "url": "http://arxiv.org/abs/2210.06005v1" }

{ "text_length": 900 }

{ "abstract": "Machine Learning (ML) has achieved enormous success in solving a variety of\nproblems in computer vision, speech recognition, object detection, to name a\nfew. The principal reason for this success is the availability of huge datasets\nfor training deep neural networks (DNNs). However, datasets can not be publicly\nreleased if they contain sensitive information such as medical or financial\nrecords. In such cases, data privacy becomes a major concern. Encryption\nmethods offer a possible solution to this issue, however their deployment on ML\napplications is non-trivial, as they seriously impact the classification\naccuracy and result in substantial computational overhead.Alternatively,\nobfuscation techniques can be used, but maintaining a good balance between\nvisual privacy and accuracy is challenging. In this work, we propose a method\nto generate secure synthetic datasets from the original private datasets. In\nour method, given a network with Batch Normalization (BN) layers pre-trained on\nthe original dataset, we first record the layer-wise BN statistics. Next, using\nthe BN statistics and the pre-trained model, we generate the synthetic dataset\nby optimizing random noises such that the synthetic data match the layer-wise\nstatistical distribution of the original model. We evaluate our method on image\nclassification dataset (CIFAR10) and show that our synthetic data can be used\nfor training networks from scratch, producing reasonable classification\nperformance.", "title": "Synthetic Dataset Generation for Privacy-Preserving Machine Learning", "url": "http://arxiv.org/abs/2210.03205v5" }

{ "text_length": 1580 }

{ "abstract": "Training of Machine Learning (ML) models in real contexts often deals with\nbig data sets and high-class imbalance samples where the class of interest is\nunrepresented (minority class). Practical solutions using classical ML models\naddress the problem of large data sets using parallel/distributed\nimplementations of training algorithms, approximate model-based solutions, or\napplying instance selection (IS) algorithms to eliminate redundant information.\nHowever, the combined problem of big and high imbalanced datasets has been less\naddressed. This work proposes three new methods for IS to be able to deal with\nlarge and imbalanced data sets. The proposed methods use Locality Sensitive\nHashing (LSH) as a base clustering technique, and then three different sampling\nmethods are applied on top of the clusters (or buckets) generated by LSH. The\nalgorithms were developed in the Apache Spark framework, guaranteeing their\nscalability. The experiments carried out in three different datasets suggest\nthat the proposed IS methods can improve the performance of a base ML model\nbetween 5% and 19% in terms of the geometric mean.", "title": "An Instance Selection Algorithm for Big Data in High imbalanced datasets based on LSH", "url": "http://arxiv.org/abs/2210.04310v1" }

{ "text_length": 1247 }

{ "abstract": "Classifying samples in incomplete datasets is a common aim for machine\nlearning practitioners, but is non-trivial. Missing data is found in most\nreal-world datasets and these missing values are typically imputed using\nestablished methods, followed by classification of the now complete, imputed,\nsamples. The focus of the machine learning researcher is then to optimise the\ndownstream classification performance. In this study, we highlight that it is\nimperative to consider the quality of the imputation. We demonstrate how the\ncommonly used measures for assessing quality are flawed and propose a new class\nof discrepancy scores which focus on how well the method recreates the overall\ndistribution of the data. To conclude, we highlight the compromised\ninterpretability of classifier models trained using poorly imputed data.", "title": "Classification of datasets with imputed missing values: does imputation quality matter?", "url": "http://arxiv.org/abs/2206.08478v1" }

{ "text_length": 950 }

{ "abstract": "While building machine learning models, Feature selection (FS) stands out as\nan essential preprocessing step used to handle the uncertainty and vagueness in\nthe data. Recently, the minimum Redundancy and Maximum Relevance (mRMR)\napproach has proven to be effective in obtaining the irredundant feature\nsubset. Owing to the generation of voluminous datasets, it is essential to\ndesign scalable solutions using distributed/parallel paradigms. MapReduce\nsolutions are proven to be one of the best approaches to designing\nfault-tolerant and scalable solutions. This work analyses the existing\nMapReduce approaches for mRMR feature selection and identifies the limitations\nthereof. In the current study, we proposed VMR_mRMR, an efficient vertical\npartitioning-based approach using a memorization approach, thereby overcoming\nthe extant approaches limitations. The experiment analysis says that VMR_mRMR\nsignificantly outperformed extant approaches and achieved a better\ncomputational gain (C.G). In addition, we also conducted a comparative analysis\nwith the horizontal partitioning approach HMR_mRMR [1] to assess the strengths\nand limitations of the proposed approach.", "title": "Scalable mRMR feature selection to handle high dimensional datasets: Vertical partitioning based Iterative MapReduce framework", "url": "http://arxiv.org/abs/2208.09901v1" }

{ "text_length": 1327 }

{ "abstract": "Estimating depth from RGB images can facilitate many computer vision tasks,\nsuch as indoor localization, height estimation, and simultaneous localization\nand mapping (SLAM). Recently, monocular depth estimation has obtained great\nprogress owing to the rapid development of deep learning techniques. They\nsurpass traditional machine learning-based methods by a large margin in terms\nof accuracy and speed. Despite the rapid progress in this topic, there are\nlacking of a comprehensive review, which is needed to summarize the current\nprogress and provide the future directions. In this survey, we first introduce\nthe datasets for depth estimation, and then give a comprehensive introduction\nof the methods from three perspectives: supervised learning-based methods,\nunsupervised learning-based methods, and sparse samples guidance-based methods.\nIn addition, downstream applications that benefit from the progress have also\nbeen illustrated. Finally, we point out the future directions and conclude the\npaper.", "title": "Deep Learning based Monocular Depth Prediction: Datasets, Methods and Applications", "url": "http://arxiv.org/abs/2011.04123v1" }

{ "text_length": 1125 }

{ "abstract": "Many recent neural models have shown remarkable empirical results in Machine\nReading Comprehension, but evidence suggests sometimes the models take\nadvantage of dataset biases to predict and fail to generalize on out-of-sample\ndata. While many other approaches have been proposed to address this issue from\nthe computation perspective such as new architectures or training procedures,\nwe believe a method that allows researchers to discover biases, and adjust the\ndata or the models in an earlier stage will be beneficial. Thus, we introduce\nMRCLens, a toolkit that detects whether biases exist before users train the\nfull model. For the convenience of introducing the toolkit, we also provide a\ncategorization of common biases in MRC.", "title": "MRCLens: an MRC Dataset Bias Detection Toolkit", "url": "http://arxiv.org/abs/2207.08943v1" }

{ "text_length": 816 }

{ "abstract": "Data owners face increasing liability for how the use of their data could\nharm under-priviliged communities. Stakeholders would like to identify the\ncharacteristics of data that lead to algorithms being biased against any\nparticular demographic groups, for example, defined by their race, gender, age,\nand/or religion. Specifically, we are interested in identifying subsets of the\nfeature space where the ground truth response function from features to\nobserved outcomes differs across demographic groups. To this end, we propose\nFORESEE, a FORESt of decision trEEs algorithm, which generates a score that\ncaptures how likely an individual's response varies with sensitive attributes.\nEmpirically, we find that our approach allows us to identify the individuals\nwho are most likely to be misclassified by several classifiers, including\nRandom Forest, Logistic Regression, Support Vector Machine, and k-Nearest\nNeighbors. The advantage of our approach is that it allows stakeholders to\ncharacterize risky samples that may contribute to discrimination, as well as,\nuse the FORESEE to estimate the risk of upcoming samples.", "title": "Ex-Ante Assessment of Discrimination in Dataset", "url": "http://arxiv.org/abs/2208.07918v2" }

{ "text_length": 1202 }

{ "abstract": "Vision-and-language tasks have increasingly drawn more attention as a means\nto evaluate human-like reasoning in machine learning models. A popular task in\nthe field is visual question answering (VQA), which aims to answer questions\nabout images. However, VQA models have been shown to exploit language bias by\nlearning the statistical correlations between questions and answers without\nlooking into the image content: e.g., questions about the color of a banana are\nanswered with yellow, even if the banana in the image is green. If societal\nbias (e.g., sexism, racism, ableism, etc.) is present in the training data,\nthis problem may be causing VQA models to learn harmful stereotypes. For this\nreason, we investigate gender and racial bias in five VQA datasets. In our\nanalysis, we find that the distribution of answers is highly different between\nquestions about women and men, as well as the existence of detrimental\ngender-stereotypical samples. Likewise, we identify that specific race-related\nattributes are underrepresented, whereas potentially discriminatory samples\nappear in the analyzed datasets. Our findings suggest that there are dangers\nassociated to using VQA datasets without considering and dealing with the\npotentially harmful stereotypes. We conclude the paper by proposing solutions\nto alleviate the problem before, during, and after the dataset collection\nprocess.", "title": "Gender and Racial Bias in Visual Question Answering Datasets", "url": "http://arxiv.org/abs/2205.08148v3" }

{ "text_length": 1482 }

{ "abstract": "The area affected by the earthquake is vast and often difficult to entirely\ncover, and the earthquake itself is a sudden event that causes multiple defects\nsimultaneously, that cannot be effectively traced using traditional, manual\nmethods. This article presents an innovative approach to the problem of\ndetecting damage after sudden events by using an interconnected set of deep\nmachine learning models organized in a single pipeline and allowing for easy\nmodification and swapping models seamlessly. Models in the pipeline were\ntrained with a synthetic dataset and were adapted to be further evaluated and\nused with unmanned aerial vehicles (UAVs) in real-world conditions. Thanks to\nthe methods presented in the article, it is possible to obtain high accuracy in\ndetecting buildings defects, segmenting constructions into their components and\nestimating their technical condition based on a single drone flight.", "title": "Computer Vision based inspection on post-earthquake with UAV synthetic dataset", "url": "http://arxiv.org/abs/2210.05282v1" }

{ "text_length": 1027 }

{ "abstract": "Commercial ML APIs offered by providers such as Google, Amazon and Microsoft\nhave dramatically simplified ML adoption in many applications. Numerous\ncompanies and academics pay to use ML APIs for tasks such as object detection,\nOCR and sentiment analysis. Different ML APIs tackling the same task can have\nvery heterogeneous performance. Moreover, the ML models underlying the APIs\nalso evolve over time. As ML APIs rapidly become a valuable marketplace and a\nwidespread way to consume machine learning, it is critical to systematically\nstudy and compare different APIs with each other and to characterize how APIs\nchange over time. However, this topic is currently underexplored due to the\nlack of data. In this paper, we present HAPI (History of APIs), a longitudinal\ndataset of 1,761,417 instances of commercial ML API applications (involving\nAPIs from Amazon, Google, IBM, Microsoft and other providers) across diverse\ntasks including image tagging, speech recognition and text mining from 2020 to\n2022. Each instance consists of a query input for an API (e.g., an image or\ntext) along with the API's output prediction/annotation and confidence scores.\nHAPI is the first large-scale dataset of ML API usages and is a unique resource\nfor studying ML-as-a-service (MLaaS). As examples of the types of analyses that\nHAPI enables, we show that ML APIs' performance change substantially over\ntime--several APIs' accuracies dropped on specific benchmark datasets. Even\nwhen the API's aggregate performance stays steady, its error modes can shift\nacross different subtypes of data between 2020 and 2022. Such changes can\nsubstantially impact the entire analytics pipelines that use some ML API as a\ncomponent. We further use HAPI to study commercial APIs' performance\ndisparities across demographic subgroups over time. HAPI can stimulate more\nresearch in the growing field of MLaaS.", "title": "HAPI: A Large-scale Longitudinal Dataset of Commercial ML API Predictions", "url": "http://arxiv.org/abs/2209.08443v1" }

{ "text_length": 1988 }

{ "abstract": "Real-time machine learning detection algorithms are often found within\nautonomous vehicle technology and depend on quality datasets. It is essential\nthat these algorithms work correctly in everyday conditions as well as under\nstrong sun glare. Reports indicate glare is one of the two most prominent\nenvironment-related reasons for crashes. However, existing datasets, such as\nLISA and the German Traffic Sign Recognition Benchmark, do not reflect the\nexistence of sun glare at all. This paper presents the GLARE traffic sign\ndataset: a collection of images with U.S based traffic signs under heavy visual\ninterference by sunlight. GLARE contains 2,157 images of traffic signs with sun\nglare, pulled from 33 videos of dashcam footage of roads in the United States.\nIt provides an essential enrichment to the widely used LISA Traffic Sign\ndataset. Our experimental study shows that although several state-of-the-art\nbaseline methods demonstrate superior performance when trained and tested\nagainst traffic sign datasets without sun glare, they greatly suffer when\ntested against GLARE (e.g., ranging from 9% to 21% mean mAP, which is\nsignificantly lower than the performances on LISA dataset). We also notice that\ncurrent architectures have better detection accuracy (e.g., on average 42% mean\nmAP gain for mainstream algorithms) when trained on images of traffic signs in\nsun glare.", "title": "GLARE: A Dataset for Traffic Sign Detection in Sun Glare", "url": "http://arxiv.org/abs/2209.08716v1" }

{ "text_length": 1473 }

{ "abstract": "Computed tomography (CT) has been widely explored as a COVID-19 screening and\nassessment tool to complement RT-PCR testing. To assist radiologists with\nCT-based COVID-19 screening, a number of computer-aided systems have been\nproposed. However, many proposed systems are built using CT data which is\nlimited in both quantity and diversity. Motivated to support efforts in the\ndevelopment of machine learning-driven screening systems, we introduce COVIDx\nCT-3, a large-scale multinational benchmark dataset for detection of COVID-19\ncases from chest CT images. COVIDx CT-3 includes 431,205 CT slices from 6,068\npatients across at least 17 countries, which to the best of our knowledge\nrepresents the largest, most diverse dataset of COVID-19 CT images in\nopen-access form. Additionally, we examine the data diversity and potential\nbiases of the COVIDx CT-3 dataset, finding that significant geographic and\nclass imbalances remain despite efforts to curate data from a wide variety of\nsources.", "title": "COVIDx CT-3: A Large-scale, Multinational, Open-Source Benchmark Dataset for Computer-aided COVID-19 Screening from Chest CT Images", "url": "http://arxiv.org/abs/2206.03043v3" }

{ "text_length": 1157 }

{ "abstract": "Distributional shift, or the mismatch between training and deployment data,\nis a significant obstacle to the usage of machine learning in high-stakes\nindustrial applications, such as autonomous driving and medicine. This creates\na need to be able to assess how robustly ML models generalize as well as the\nquality of their uncertainty estimates. Standard ML baseline datasets do not\nallow these properties to be assessed, as the training, validation and test\ndata are often identically distributed. Recently, a range of dedicated\nbenchmarks have appeared, featuring both distributionally matched and shifted\ndata. Among these benchmarks, the Shifts dataset stands out in terms of the\ndiversity of tasks as well as the data modalities it features. While most of\nthe benchmarks are heavily dominated by 2D image classification tasks, Shifts\ncontains tabular weather forecasting, machine translation, and vehicle motion\nprediction tasks. This enables the robustness properties of models to be\nassessed on a diverse set of industrial-scale tasks and either universal or\ndirectly applicable task-specific conclusions to be reached. In this paper, we\nextend the Shifts Dataset with two datasets sourced from industrial, high-risk\napplications of high societal importance. Specifically, we consider the tasks\nof segmentation of white matter Multiple Sclerosis lesions in 3D magnetic\nresonance brain images and the estimation of power consumption in marine cargo\nvessels. Both tasks feature ubiquitous distributional shifts and a strict\nsafety requirement due to the high cost of errors. These new datasets will\nallow researchers to further explore robust generalization and uncertainty\nestimation in new situations. In this work, we provide a description of the\ndataset and baseline results for both tasks.", "title": "Shifts 2.0: Extending The Dataset of Real Distributional Shifts", "url": "http://arxiv.org/abs/2206.15407v2" }

{ "text_length": 1897 }

{ "abstract": "Cerebral stroke, the second most substantial cause of death universally, has\nbeen a primary public health concern over the last few years. With the help of\nmachine learning techniques, early detection of various stroke alerts is\naccessible, which can efficiently prevent or diminish the stroke. Medical\ndataset, however, are frequently unbalanced in their class label, with a\ntendency to poorly predict minority classes. In this paper, the potential risk\nfactors for stroke are investigated. Moreover, four distinctive approaches are\napplied to improve the classification of the minority class in the imbalanced\nstroke dataset, which are the ensemble weight voting classifier, the Synthetic\nMinority Over-sampling Technique (SMOTE), Principal Component Analysis with\nK-Means Clustering (PCA-Kmeans), Focal Loss with the Deep Neural Network (DNN)\nand compare their performance. Through the analysis results, SMOTE and\nPCA-Kmeans with DNN-Focal Loss work best for the limited size of a large severe\nimbalanced dataset,which is 2-4 times outperform Kaggle work.", "title": "Machine Learning Performance Analysis to Predict Stroke Based on Imbalanced Medical Dataset", "url": "http://arxiv.org/abs/2211.07652v1" }

{ "text_length": 1184 }

{ "abstract": "Various problems of any credit card fraud detection based on machine learning\ncome from the imbalanced aspect of transaction datasets. Indeed, the number of\nfrauds compared to the number of regular transactions is tiny and has been\nshown to damage learning performances, e.g., at worst, the algorithm can learn\nto classify all the transactions as regular. Resampling methods and\ncost-sensitive approaches are known to be good candidates to leverage this\nissue of imbalanced datasets. This paper evaluates numerous state-of-the-art\nresampling methods on a large real-life online credit card payments dataset. We\nshow they are inefficient because methods are intractable or because metrics do\nnot exhibit substantial improvements. Our work contributes to this domain in\n(1) that we compare many state-of-the-art resampling methods on a large-scale\ndataset and in (2) that we use a real-life online credit card payments dataset.", "title": "Evaluating resampling methods on a real-life highly imbalanced online credit card payments dataset", "url": "http://arxiv.org/abs/2206.13152v1" }

{ "text_length": 1058 }

{ "abstract": "The conversation scenario is one of the most important and most challenging\nscenarios for speech processing technologies because people in conversation\nrespond to each other in a casual style. Detecting the speech activities of\neach person in a conversation is vital to downstream tasks, like natural\nlanguage processing, machine translation, etc. People refer to the detection\ntechnology of \"who speak when\" as speaker diarization (SD). Traditionally,\ndiarization error rate (DER) has been used as the standard evaluation metric of\nSD systems for a long time. However, DER fails to give enough importance to\nshort conversational phrases, which are short but important on the semantic\nlevel. Also, a carefully and accurately manually-annotated testing dataset\nsuitable for evaluating the conversational SD technologies is still unavailable\nin the speech community. In this paper, we design and describe the\nConversational Short-phrases Speaker Diarization (CSSD) task, which consists of\ntraining and testing datasets, evaluation metric and baselines. In the dataset\naspect, despite the previously open-sourced 180-hour conversational\nMagicData-RAMC dataset, we prepare an individual 20-hour conversational speech\ntest dataset with carefully and artificially verified speakers timestamps\nannotations for the CSSD task. In the metric aspect, we design the new\nconversational DER (CDER) evaluation metric, which calculates the SD accuracy\nat the utterance level. In the baseline aspect, we adopt a commonly used\nmethod: Variational Bayes HMM x-vector system, as the baseline of the CSSD\ntask. Our evaluation metric is publicly available at\nhttps://github.com/SpeechClub/CDER_Metric.", "title": "The Conversational Short-phrase Speaker Diarization (CSSD) Task: Dataset, Evaluation Metric and Baselines", "url": "http://arxiv.org/abs/2208.08042v1" }

{ "text_length": 1819 }

{ "abstract": "In this dataset we provide a comprehensive collection of magnetograms (images\nquantifying the strength of the magnetic field) from the National Aeronautics\nand Space Administration's (NASA's) Solar Dynamics Observatory (SDO). The\ndataset incorporates data from three sources and provides SDO Helioseismic and\nMagnetic Imager (HMI) magnetograms of solar active regions (regions of large\nmagnetic flux, generally the source of eruptive events) as well as labels of\ncorresponding flaring activity. This dataset will be useful for image analysis\nor solar physics research related to magnetic structure, its evolution over\ntime, and its relation to solar flares. The dataset will be of interest to\nthose researchers investigating automated solar flare prediction methods,\nincluding supervised and unsupervised machine learning (classical and deep),\nbinary and multi-class classification, and regression. This dataset is a\nminimally processed, user configurable dataset of consistently sized images of\nsolar active regions that can serve as a benchmark dataset for solar flare\nprediction research.", "title": "Solar Active Region Magnetogram Image Dataset for Studies of Space Weather", "url": "http://arxiv.org/abs/2305.09492v2" }

{ "text_length": 1200 }

{ "abstract": "Effective Human-AI teaming requires the ability to communicate the goals of\nthe team and constraints under which you need the agent to operate. Providing\nthe ability to specify the shared intent or operation criteria of the team can\nenable an AI agent to perform its primary function while still being able to\ncater to the specific desires of the current team. While significant work has\nbeen conducted to instruct an agent to perform a task, via language or\ndemonstrations, prior work lacks a focus on building agents which can operate\nwithin the parameters specified by a team. Worse yet, there is a dearth of\nresearch pertaining to enabling humans to provide their specifications through\nunstructured, naturalist language. In this paper, we propose the use of goals\nand constraints as a scaffold to modulate and evaluate autonomous agents. We\ncontribute to this field by presenting a novel dataset, and an associated data\ncollection protocol, which maps language descriptions to goals and constraints\ncorresponding to specific strategies developed by human participants for the\nboard game Risk. Leveraging state-of-the-art language models and augmentation\nprocedures, we develop a machine learning framework which can be used to\nidentify goals and constraints from unstructured strategy descriptions. To\nempirically validate our approach we conduct a human-subjects study to\nestablish a human-baseline for our dataset. Our results show that our machine\nlearning architecture is better able to interpret unstructured language\ndescriptions into strategy specifications than human raters tasked with\nperforming the same machine translation task (F(1,272.53) = 17.025, p < 0.001).", "title": "Commander's Intent: A Dataset and Modeling Approach for Human-AI Task Specification in Strategic Play", "url": "http://arxiv.org/abs/2208.08374v1" }

{ "text_length": 1815 }

{ "abstract": "The growing use of supervised machine learning in research and industry has\nincreased the need for labeled datasets. Crowdsourcing has emerged as a popular\nmethod to create data labels. However, working on large batches of tasks leads\nto worker fatigue, negatively impacting labeling quality. To address this, we\npresent TruEyes, a collaborative crowdsourcing system, enabling the\ndistribution of micro-tasks to mobile app users. TruEyes allows machine\nlearning practitioners to publish labeling tasks, mobile app developers to\nintegrate task ads for monetization, and users to label data instead of\nwatching advertisements. To evaluate the system, we conducted an experiment\nwith N=296 participants. Our results show that the quality of the labeled data\nis comparable to traditional crowdsourcing approaches and most users prefer\ntask ads over traditional ads. We discuss extensions to the system and address\nhow mobile advertisement space can be used as a productive resource in the\nfuture.", "title": "TruEyes: Utilizing Microtasks in Mobile Apps for Crowdsourced Labeling of Machine Learning Datasets", "url": "http://arxiv.org/abs/2209.14708v1" }

{ "text_length": 1126 }

{ "abstract": "Camera images are ubiquitous in machine learning research. They also play a\ncentral role in the delivery of important services spanning medicine and\nenvironmental surveying. However, the application of machine learning models in\nthese domains has been limited because of robustness concerns. A primary\nfailure mode are performance drops due to differences between the training and\ndeployment data. While there are methods to prospectively validate the\nrobustness of machine learning models to such dataset drifts, existing\napproaches do not account for explicit models of the primary object of\ninterest: the data. This limits our ability to study and understand the\nrelationship between data generation and downstream machine learning model\nperformance in a physically accurate manner. In this study, we demonstrate how\nto overcome this limitation by pairing traditional machine learning with\nphysical optics to obtain explicit and differentiable data models. We\ndemonstrate how such data models can be constructed for image data and used to\ncontrol downstream machine learning model performance related to dataset drift.\nThe findings are distilled into three applications. First, drift synthesis\nenables the controlled generation of physically faithful drift test cases to\npower model selection and targeted generalization. Second, the gradient\nconnection between machine learning task model and data model allows advanced,\nprecise tolerancing of task model sensitivity to changes in the data\ngeneration. These drift forensics can be used to precisely specify the\nacceptable data environments in which a task model may be run. Third, drift\noptimization opens up the possibility to create drifts that can help the task\nmodel learn better faster, effectively optimizing the data generating process\nitself. A guide to access the open code and datasets is available at\nhttps://github.com/aiaudit-org/raw2logit.", "title": "Data Models for Dataset Drift Controls in Machine Learning With Optical Images", "url": "http://arxiv.org/abs/2211.02578v3" }

{ "text_length": 2020 }

{ "abstract": "This article documents a dataset of sentence pairs between Faroese and\nDanish, produced at ITU Copenhagen. The data covers tranlsation from both\nsource languages, and is intended for use as training data for machine\ntranslation systems in this language pair.", "title": "The ITU Faroese Pairs Dataset", "url": "http://arxiv.org/abs/2206.08727v1" }

{ "text_length": 322 }

{ "abstract": "Machine learning algorithms have been widely used in intrusion detection\nsystems, including Multi-layer Perceptron (MLP). In this study, we proposed a\ntwo-stage model that combines the Birch clustering algorithm and MLP classifier\nto improve the performance of network anomaly multi-classification. In our\nproposed method, we first apply Birch or Kmeans as an unsupervised clustering\nalgorithm to the CICIDS-2017 dataset to pre-group the data. The generated\npseudo-label is then added as an additional feature to the training of the\nMLP-based classifier. The experimental results show that using Birch and\nK-Means clustering for data pre-grouping can improve intrusion detection system\nperformance. Our method can achieve 99.73% accuracy in multi-classification\nusing Birch clustering, which is better than similar researches using a\nstand-alone MLP model.", "title": "Improving Multilayer-Perceptron(MLP)-based Network Anomaly Detection with Birch Clustering on CICIDS-2017 Dataset", "url": "http://arxiv.org/abs/2208.09711v2" }

{ "text_length": 1004 }

{ "abstract": "Scientific action graphs extraction from materials synthesis procedures is\nimportant for reproducible research, machine automation, and material\nprediction. But the lack of annotated data has hindered progress in this field.\nWe demonstrate an effort to annotate Polycrystalline Materials Synthesis\nProcedures (PcMSP) from 305 open access scientific articles for the\nconstruction of synthesis action graphs. This is a new dataset for material\nscience information extraction that simultaneously contains the synthesis\nsentences extracted from the experimental paragraphs, as well as the entity\nmentions and intra-sentence relations. A two-step human annotation and\ninter-annotator agreement study guarantee the high quality of the PcMSP corpus.\nWe introduce four natural language processing tasks: sentence classification,\nnamed entity recognition, relation classification, and joint extraction of\nentities and relations. Comprehensive experiments validate the effectiveness of\nseveral state-of-the-art models for these challenges while leaving large space\nfor improvement. We also perform the error analysis and point out some unique\nchallenges that require further investigation. We will release our annotation\nscheme, the corpus, and codes to the research community to alleviate the\nscarcity of labeled data in this domain.", "title": "PcMSP: A Dataset for Scientific Action Graphs Extraction from Polycrystalline Materials Synthesis Procedure Text", "url": "http://arxiv.org/abs/2210.12401v1" }

{ "text_length": 1471 }

{ "abstract": "The development of semi-supervised learning techniques is essential to\nenhance the generalization capacities of machine learning algorithms. Indeed,\nraw image data are abundant while labels are scarce, therefore it is crucial to\nleverage unlabeled inputs to build better models. The availability of large\ndatabases have been key for the development of learning algorithms with high\nlevel performance.\n Despite the major role of machine learning in Earth Observation to derive\nproducts such as land cover maps, datasets in the field are still limited,\neither because of modest surface coverage, lack of variety of scenes or\nrestricted classes to identify. We introduce a novel large-scale dataset for\nsemi-supervised semantic segmentation in Earth Observation, the MiniFrance\nsuite. MiniFrance has several unprecedented properties: it is large-scale,\ncontaining over 2000 very high resolution aerial images, accounting for more\nthan 200 billions samples (pixels); it is varied, covering 16 conurbations in\nFrance, with various climates, different landscapes, and urban as well as\ncountryside scenes; and it is challenging, considering land use classes with\nhigh-level semantics. Nevertheless, the most distinctive quality of MiniFrance\nis being the only dataset in the field especially designed for semi-supervised\nlearning: it contains labeled and unlabeled images in its training partition,\nwhich reproduces a life-like scenario. Along with this dataset, we present\ntools for data representativeness analysis in terms of appearance similarity\nand a thorough study of MiniFrance data, demonstrating that it is suitable for\nlearning and generalizes well in a semi-supervised setting. Finally, we present\nsemi-supervised deep architectures based on multi-task learning and the first\nexperiments on MiniFrance.", "title": "Semi-Supervised Semantic Segmentation in Earth Observation: The MiniFrance Suite, Dataset Analysis and Multi-task Network Study", "url": "http://arxiv.org/abs/2010.07830v1" }

{ "text_length": 1970 }

{ "abstract": "We introduce RaidaR, a rich annotated image dataset of rainy street scenes,\nto support autonomous driving research. The new dataset contains the largest\nnumber of rainy images (58,542) to date, 5,000 of which provide semantic\nsegmentations and 3,658 provide object instance segmentations. The RaidaR\nimages cover a wide range of realistic rain-induced artifacts, including fog,\ndroplets, and road reflections, which can effectively augment existing street\nscene datasets to improve data-driven machine perception during rainy weather.\nTo facilitate efficient annotation of a large volume of images, we develop a\nsemi-automatic scheme combining manual segmentation and an automated processing\nakin to cross validation, resulting in 10-20 fold reduction on annotation time.\nWe demonstrate the utility of our new dataset by showing how data augmentation\nwith RaidaR can elevate the accuracy of existing segmentation algorithms. We\nalso present a novel unpaired image-to-image translation algorithm for\nadding/removing rain artifacts, which directly benefits from RaidaR.", "title": "RaidaR: A Rich Annotated Image Dataset of Rainy Street Scenes", "url": "http://arxiv.org/abs/2104.04606v3" }

{ "text_length": 1163 }

{ "abstract": "To better interact with users, a social robot should understand the users'\nbehavior, infer the intention, and respond appropriately. Machine learning is\none way of implementing robot intelligence. It provides the ability to\nautomatically learn and improve from experience instead of explicitly telling\nthe robot what to do. Social skills can also be learned through watching\nhuman-human interaction videos. However, human-human interaction datasets are\nrelatively scarce to learn interactions that occur in various situations.\nMoreover, we aim to use service robots in the elderly-care domain; however,\nthere has been no interaction dataset collected for this domain. For this\nreason, we introduce a human-human interaction dataset for teaching non-verbal\nsocial behaviors to robots. It is the only interaction dataset that elderly\npeople have participated in as performers. We recruited 100 elderly people and\ntwo college students to perform 10 interactions in an indoor environment. The\nentire dataset has 5,000 interaction samples, each of which contains depth\nmaps, body indexes and 3D skeletal data that are captured with three Microsoft\nKinect v2 cameras. In addition, we provide the joint angles of a humanoid NAO\nrobot which are converted from the human behavior that robots need to learn.\nThe dataset and useful python scripts are available for download at\nhttps://github.com/ai4r/AIR-Act2Act. It can be used to not only teach social\nskills to robots but also benchmark action recognition algorithms.", "title": "AIR-Act2Act: Human-human interaction dataset for teaching non-verbal social behaviors to robots", "url": "http://arxiv.org/abs/2009.02041v1" }

{ "text_length": 1639 }

{ "abstract": "Deep neural networks (DNNs) have been shown to outperform traditional machine\nlearning algorithms in a broad variety of application domains due to their\neffectiveness in modeling complex problems and handling high-dimensional\ndatasets. Many real-life datasets, however, are of increasingly high\ndimensionality, where a large number of features may be irrelevant for both\nsupervised and unsupervised learning tasks. The inclusion of such features\nwould not only introduce unwanted noise but also increase computational\ncomplexity. Furthermore, due to high non-linearity and dependency among a large\nnumber of features, DNN models tend to be unavoidably opaque and perceived as\nblack-box methods because of their not well-understood internal functioning.\nTheir algorithmic complexity is often simply beyond the capacities of humans to\nunderstand the interplay among myriads of hyperparameters. A well-interpretable\nmodel can identify statistically significant features and explain the way they\naffect the model's outcome. In this paper, we propose an efficient method to\nimprove the interpretability of black-box models for classification tasks in\nthe case of high-dimensional datasets. First, we train a black-box model on a\nhigh-dimensional dataset to learn the embeddings on which the classification is\nperformed. To decompose the inner working principles of the black-box model and\nto identify top-k important features, we employ different probing and\nperturbing techniques. We then approximate the behavior of the black-box model\nby means of an interpretable surrogate model on the top-k feature space.\nFinally, we derive decision rules and local explanations from the surrogate\nmodel to explain individual decisions. Our approach outperforms\nstate-of-the-art methods like TabNet and XGboost when tested on different\ndatasets with varying dimensionality between 50 and 20,000 w.r.t metrics and\nexplainability.", "title": "Interpreting Black-box Machine Learning Models for High Dimensional Datasets", "url": "http://arxiv.org/abs/2208.13405v2" }

{ "text_length": 2023 }

{ "abstract": "Recent advances in machine learning have significantly improved the\nunderstanding of source code data and achieved good performance on a number of\ndownstream tasks. Open source repositories like GitHub enable this process with\nrich unlabeled code data. However, the lack of high quality labeled data has\nlargely hindered the progress of several code related tasks, such as program\ntranslation, summarization, synthesis, and code search. This paper introduces\nXLCoST, Cross-Lingual Code SnippeT dataset, a new benchmark dataset for\ncross-lingual code intelligence. Our dataset contains fine-grained parallel\ndata from 8 languages (7 commonly used programming languages and English), and\nsupports 10 cross-lingual code tasks. To the best of our knowledge, it is the\nlargest parallel dataset for source code both in terms of size and the number\nof languages. We also provide the performance of several state-of-the-art\nbaseline models for each task. We believe this new dataset can be a valuable\nasset for the research community and facilitate the development and validation\nof new methods for cross-lingual code intelligence.", "title": "XLCoST: A Benchmark Dataset for Cross-lingual Code Intelligence", "url": "http://arxiv.org/abs/2206.08474v1" }

{ "text_length": 1221 }

{ "abstract": "Enhancing interoperability and information exchange between domain-specific\nsoftware products for BIM is an important aspect in the Architecture,\nEngineering, Construction and Operations industry. Recent research started\ninvestigating methods from the areas of machine and deep learning for semantic\nenrichment of BIM models. However, training and evaluation of these machine\nlearning algorithms requires sufficiently large and comprehensive datasets.\nThis work presents IFCNet, a dataset of single-entity IFC files spanning a\nbroad range of IFC classes containing both geometric and semantic information.\nUsing only the geometric information of objects, the experiments show that\nthree different deep learning models are able to achieve good classification\nperformance.", "title": "IFCNet: A Benchmark Dataset for IFC Entity Classification", "url": "http://arxiv.org/abs/2106.09712v1" }

{ "text_length": 862 }

{ "abstract": "With the growing rates of cyber-attacks and cyber espionage, the need for\nbetter and more powerful intrusion detection systems (IDS) is even more\nwarranted nowadays. The basic task of an IDS is to act as the first line of\ndefense, in detecting attacks on the internet. As intrusion tactics from\nintruders become more sophisticated and difficult to detect, researchers have\nstarted to apply novel Machine Learning (ML) techniques to effectively detect\nintruders and hence preserve internet users' information and overall trust in\nthe entire internet network security. Over the last decade, there has been an\nexplosion of research on intrusion detection techniques based on ML and Deep\nLearning (DL) architectures on various cyber security-based datasets such as\nthe DARPA, KDDCUP'99, NSL-KDD, CAIDA, CTU-13, UNSW-NB15. In this research, we\nreview contemporary literature and provide a comprehensive survey of different\ntypes of intrusion detection technique that applies Support Vector Machines\n(SVMs) algorithms as a classifier. We focus only on studies that have been\nevaluated on the two most widely used datasets in cybersecurity namely: the\nKDDCUP'99 and the NSL-KDD datasets. We provide a summary of each method,\nidentifying the role of the SVMs classifier, and all other algorithms involved\nin the studies. Furthermore, we present a critical review of each method, in\ntabular form, highlighting the performance measures, strengths, and limitations\nof each of the methods surveyed.", "title": "Intrusion Detection Systems Using Support Vector Machines on the KDDCUP'99 and NSL-KDD Datasets: A Comprehensive Survey", "url": "http://arxiv.org/abs/2209.05579v1" }

{ "text_length": 1640 }

{ "abstract": "This paper describes NEREL-BIO -- an annotation scheme and corpus of PubMed\nabstracts in Russian and smaller number of abstracts in English. NEREL-BIO\nextends the general domain dataset NEREL by introducing domain-specific entity\ntypes. NEREL-BIO annotation scheme covers both general and biomedical domains\nmaking it suitable for domain transfer experiments. NEREL-BIO provides\nannotation for nested named entities as an extension of the scheme employed for\nNEREL. Nested named entities may cross entity boundaries to connect to shorter\nentities nested within longer entities, making them harder to detect.\n NEREL-BIO contains annotations for 700+ Russian and 100+ English abstracts.\nAll English PubMed annotations have corresponding Russian counterparts. Thus,\nNEREL-BIO comprises the following specific features: annotation of nested named\nentities, it can be used as a benchmark for cross-domain (NEREL -> NEREL-BIO)\nand cross-language (English -> Russian) transfer. We experiment with both\ntransformer-based sequence models and machine reading comprehension (MRC)\nmodels and report their results.\n The dataset is freely available at https://github.com/nerel-ds/NEREL-BIO.", "title": "NEREL-BIO: A Dataset of Biomedical Abstracts Annotated with Nested Named Entities", "url": "http://arxiv.org/abs/2210.11913v1" }

{ "text_length": 1294 }

{ "abstract": "One of the challenges in virtual environments is the difficulty users have in\ninteracting with these increasingly complex systems. Ultimately, endowing\nmachines with the ability to perceive users emotions will enable a more\nintuitive and reliable interaction. Consequently, using the\nelectroencephalogram as a bio-signal sensor, the affective state of a user can\nbe modelled and subsequently utilised in order to achieve a system that can\nrecognise and react to the user's emotions. This paper investigates features\nextracted from electroencephalogram signals for the purpose of affective state\nmodelling based on Russell's Circumplex Model. Investigations are presented\nthat aim to provide the foundation for future work in modelling user affect to\nenhance interaction experience in virtual environments. The DEAP dataset was\nused within this work, along with a Support Vector Machine and Random Forest,\nwhich yielded reasonable classification accuracies for Valence and Arousal\nusing feature vectors based on statistical measurements and band power from the\n\\'z, \\b{eta}, \\'z, and \\'z\\'z waves and High Order Crossing of the EEG signal.", "title": "Towards emotion recognition for virtual environments: an evaluation of EEG features on benchmark dataset", "url": "http://arxiv.org/abs/2210.13876v1" }

{ "text_length": 1277 }

{ "abstract": "The COVID-19 pandemic continues to bring up various topics discussed or\ndebated on social media. In order to explore the impact of pandemics on\npeople's lives, it is crucial to understand the public's concerns and attitudes\ntowards pandemic-related entities (e.g., drugs, vaccines) on social media.\nHowever, models trained on existing named entity recognition (NER) or targeted\nsentiment analysis (TSA) datasets have limited ability to understand\nCOVID-19-related social media texts because these datasets are not designed or\nannotated from a medical perspective. This paper releases METS-CoV, a dataset\ncontaining medical entities and targeted sentiments from COVID-19-related\ntweets. METS-CoV contains 10,000 tweets with 7 types of entities, including 4\nmedical entity types (Disease, Drug, Symptom, and Vaccine) and 3 general entity\ntypes (Person, Location, and Organization). To further investigate tweet users'\nattitudes toward specific entities, 4 types of entities (Person, Organization,\nDrug, and Vaccine) are selected and annotated with user sentiments, resulting\nin a targeted sentiment dataset with 9,101 entities (in 5,278 tweets). To the\nbest of our knowledge, METS-CoV is the first dataset to collect medical\nentities and corresponding sentiments of COVID-19-related tweets. We benchmark\nthe performance of classical machine learning models and state-of-the-art deep\nlearning models on NER and TSA tasks with extensive experiments. Results show\nthat the dataset has vast room for improvement for both NER and TSA tasks.\nMETS-CoV is an important resource for developing better medical social media\ntools and facilitating computational social science research, especially in\nepidemiology. Our data, annotation guidelines, benchmark models, and source\ncode are publicly available (https://github.com/YLab-Open/METS-CoV) to ensure\nreproducibility.", "title": "METS-CoV: A Dataset of Medical Entity and Targeted Sentiment on COVID-19 Related Tweets", "url": "http://arxiv.org/abs/2209.13773v1" }

{ "text_length": 1979 }

{ "abstract": "Background: The volume of distribution at steady state (VDss) is a\nfundamental pharmacokinetics (PK) property of drugs, which measures how\neffectively a drug molecule is distributed throughout the body. Along with the\nclearance (CL), it determines the half-life and, therefore, the drug dosing\ninterval. However, the molecular data size limits the generalizability of the\nreported machine learning models. Objective: This study aims to provide a clean\nand comprehensive dataset for human VDss as the benchmarking data source,\nfostering and benefiting future predictive studies. Moreover, several\npredictive models were also built with machine learning regression algorithms.\nMethods: The dataset was curated from 13 publicly accessible data sources and\nthe DrugBank database entirely from intravenous drug administration and then\nunderwent extensive data cleaning. The molecular descriptors were calculated\nwith Mordred, and feature selection was conducted for constructing predictive\nmodels. Five machine learning methods were used to build regression models,\ngrid search was used to optimize hyperparameters, and ten-fold cross-validation\nwas used to evaluate the model. Results: An enriched dataset of VDss\n(https://github.com/da-wen-er/VDss) was constructed with 2440 molecules. Among\nthe prediction models, the LightGBM model was the most stable and had the best\ninternal prediction ability with Q2 = 0.837, R2=0.814 and for the other four\nmodels, Q2 was higher than 0.79. Conclusions: To the best of our knowledge,\nthis is the largest dataset for VDss, which can be used as the benchmark for\ncomputational studies of VDss. Moreover, the regression models reported within\nthis study can be of use for pharmacokinetic related studies.", "title": "A Benchmarking Dataset with 2440 Organic Molecules for Volume Distribution at Steady State", "url": "http://arxiv.org/abs/2211.05661v1" }

{ "text_length": 1863 }

{ "abstract": "We introduce a new dataset called SyntheticFur built specifically for machine\nlearning training. The dataset consists of ray traced synthetic fur renders\nwith corresponding rasterized input buffers and simulation data files. We\nprocedurally generated approximately 140,000 images and 15 simulations with\nHoudini. The images consist of fur groomed with different skin primitives and\nmove with various motions in a predefined set of lighting environments. We also\ndemonstrated how the dataset could be used with neural rendering to\nsignificantly improve fur graphics using inexpensive input buffers by training\na conditional generative adversarial network with perceptual loss. We hope the\navailability of such high fidelity fur renders will encourage new advances with\nneural rendering for a variety of applications.", "title": "SyntheticFur dataset for neural rendering", "url": "http://arxiv.org/abs/2105.06409v1" }

{ "text_length": 891 }

{ "abstract": "From face recognition in smartphones to automatic routing on self-driving\ncars, machine vision algorithms lie in the core of these features. These\nsystems solve image based tasks by identifying and understanding objects,\nsubsequently making decisions from these information. However, errors in\ndatasets are usually induced or even magnified in algorithms, at times\nresulting in issues such as recognising black people as gorillas and\nmisrepresenting ethnicities in search results. This paper tracks the errors in\ndatasets and their impacts, revealing that a flawed dataset could be a result\nof limited categories, incomprehensive sourcing and poor classification.", "title": "Seeing the Unseen: Errors and Bias in Visual Datasets", "url": "http://arxiv.org/abs/2211.01847v1" }

{ "text_length": 751 }

{ "abstract": "As a relatively new form of sport, esports offers unparalleled data\navailability. Despite the vast amounts of data that are generated by game\nengines, it can be challenging to extract them and verify their integrity for\nthe purposes of practical and scientific use.\n Our work aims to open esports to a broader scientific community by supplying\nraw and pre-processed files from StarCraft II esports tournaments. These files\ncan be used in statistical and machine learning modeling tasks and related to\nvarious laboratory-based measurements (e.g., behavioral tests, brain imaging).\nWe have gathered publicly available game-engine generated \"replays\" of\ntournament matches and performed data extraction and cleanup using a low-level\napplication programming interface (API) parser library.\n Additionally, we open-sourced and published all the custom tools that were\ndeveloped in the process of creating our dataset. These tools include PyTorch\nand PyTorch Lightning API abstractions to load and model the data.\n Our dataset contains replays from major and premiere StarCraft II tournaments\nsince 2016. To prepare the dataset, we processed 55 tournament \"replaypacks\"\nthat contained 17930 files with game-state information. Based on initial\ninvestigation of available StarCraft II datasets, we observed that our dataset\nis the largest publicly available source of StarCraft II esports data upon its\npublication.\n Analysis of the extracted data holds promise for further Artificial\nIntelligence (AI), Machine Learning (ML), psychological, Human-Computer\nInteraction (HCI), and sports-related studies in a variety of supervised and\nself-supervised tasks.", "title": "SC2EGSet: StarCraft II Esport Replay and Game-state Dataset", "url": "http://arxiv.org/abs/2207.03428v2" }

{ "text_length": 1745 }

{ "abstract": "We generate synthetic images with the \"Stable Diffusion\" image generation\nmodel using the Wordnet taxonomy and the definitions of concepts it contains.\nThis synthetic image database can be used as training data for data\naugmentation in machine learning applications, and it is used to investigate\nthe capabilities of the Stable Diffusion model.\n Analyses show that Stable Diffusion can produce correct images for a large\nnumber of concepts, but also a large variety of different representations. The\nresults show differences depending on the test concepts considered and problems\nwith very specific concepts. These evaluations were performed using a vision\ntransformer model for image classification.", "title": "Evaluating a Synthetic Image Dataset Generated with Stable Diffusion", "url": "http://arxiv.org/abs/2211.01777v2" }

{ "text_length": 804 }

{ "abstract": "Forest biomass is a key influence for future climate, and the world urgently\nneeds highly scalable financing schemes, such as carbon offsetting\ncertifications, to protect and restore forests. Current manual forest carbon\nstock inventory methods of measuring single trees by hand are time, labour, and\ncost-intensive and have been shown to be subjective. They can lead to\nsubstantial overestimation of the carbon stock and ultimately distrust in\nforest financing. The potential for impact and scale of leveraging advancements\nin machine learning and remote sensing technologies is promising but needs to\nbe of high quality in order to replace the current forest stock protocols for\ncertifications.\n In this paper, we present ReforesTree, a benchmark dataset of forest carbon\nstock in six agro-forestry carbon offsetting sites in Ecuador. Furthermore, we\nshow that a deep learning-based end-to-end model using individual tree\ndetection from low cost RGB-only drone imagery is accurately estimating forest\ncarbon stock within official carbon offsetting certification standards.\nAdditionally, our baseline CNN model outperforms state-of-the-art\nsatellite-based forest biomass and carbon stock estimates for this type of\nsmall-scale, tropical agro-forestry sites. We present this dataset to encourage\nmachine learning research in this area to increase accountability and\ntransparency of monitoring, verification and reporting (MVR) in carbon\noffsetting projects, as well as scaling global reforestation financing through\naccurate remote sensing.", "title": "ReforesTree: A Dataset for Estimating Tropical Forest Carbon Stock with Deep Learning and Aerial Imagery", "url": "http://arxiv.org/abs/2201.11192v1" }

{ "text_length": 1680 }

{ "abstract": "Trajectory data analysis is an essential component for highly automated\ndriving. Complex models developed with these data predict other road users'\nmovement and behavior patterns. Based on these predictions - and additional\ncontextual information such as the course of the road, (traffic) rules, and\ninteraction with other road users - the highly automated vehicle (HAV) must be\nable to reliably and safely perform the task assigned to it, e.g., moving from\npoint A to B. Ideally, the HAV moves safely through its environment, just as we\nwould expect a human driver to do. However, if unusual trajectories occur,\nso-called trajectory corner cases, a human driver can usually cope well, but an\nHAV can quickly get into trouble. In the definition of trajectory corner cases,\nwhich we provide in this work, we will consider the relevance of unusual\ntrajectories with respect to the task at hand. Based on this, we will also\npresent a taxonomy of different trajectory corner cases. The categorization of\ncorner cases into the taxonomy will be shown with examples and is done by cause\nand required data sources. To illustrate the complexity between the machine\nlearning (ML) model and the corner case cause, we present a general processing\nchain underlying the taxonomy.", "title": "Space, Time, and Interaction: A Taxonomy of Corner Cases in Trajectory Datasets for Automated Driving", "url": "http://arxiv.org/abs/2210.08885v1" }

{ "text_length": 1401 }

{ "abstract": "With advancements in computer vision taking place day by day, recently a lot\nof light is being shed on activity recognition. With the range for real-world\napplications utilizing this field of study increasing across a multitude of\nindustries such as security and healthcare, it becomes crucial for businesses\nto distinguish which machine learning methods perform better than others in the\narea. This paper strives to aid in this predicament i.e. building upon previous\nrelated work, it employs both classical and ensemble approaches on rich pose\nestimation (OpenPose) and HAR datasets. Making use of appropriate metrics to\nevaluate the performance for each model, the results show that overall, random\nforest yields the highest accuracy in classifying ADLs. Relatively all the\nmodels have excellent performance across both datasets, except for logistic\nregression and AdaBoost perform poorly in the HAR one. With the limitations of\nthis paper also discussed in the end, the scope for further research is vast,\nwhich can use this paper as a base in aims of producing better results.", "title": "Performance of different machine learning methods on activity recognition and pose estimation datasets", "url": "http://arxiv.org/abs/2210.10247v1" }

{ "text_length": 1218 }

{ "abstract": "Machine learning models deployed in healthcare systems face data drawn from\ncontinually evolving environments. However, researchers proposing such models\ntypically evaluate them in a time-agnostic manner, with train and test splits\nsampling patients throughout the entire study period. We introduce the\nEvaluation on Medical Datasets Over Time (EMDOT) framework and Python package,\nwhich evaluates the performance of a model class over time. Across five medical\ndatasets and a variety of models, we compare two training strategies: (1) using\nall historical data, and (2) using a window of the most recent data. We note\nchanges in performance over time, and identify possible explanations for these\nshocks.", "title": "Model Evaluation in Medical Datasets Over Time", "url": "http://arxiv.org/abs/2211.07165v1" }

{ "text_length": 786 }

{ "abstract": "The availability of different pre-trained semantic models enabled the quick\ndevelopment of machine learning components for downstream applications. Despite\nthe availability of abundant text data for low resource languages, only a few\nsemantic models are publicly available. Publicly available pre-trained models\nare usually built as a multilingual version of semantic models that can not fit\nwell for each language due to context variations. In this work, we introduce\ndifferent semantic models for Amharic. After we experiment with the existing\npre-trained semantic models, we trained and fine-tuned nine new different\nmodels using a monolingual text corpus. The models are build using word2Vec\nembeddings, distributional thesaurus (DT), contextual embeddings, and DT\nembeddings obtained via network embedding algorithms. Moreover, we employ these\nmodels for different NLP tasks and investigate their impact. We find that newly\ntrained models perform better than pre-trained multilingual models.\nFurthermore, models based on contextual embeddings from RoBERTA perform better\nthan the word2Vec models.", "title": "Introducing various Semantic Models for Amharic: Experimentation and Evaluation with multiple Tasks and Datasets", "url": "http://arxiv.org/abs/2011.01154v2" }

{ "text_length": 1248 }

{ "abstract": "We introduce QM7-X, a comprehensive dataset of 42 physicochemical properties\nfor $\\approx$ 4.2 M equilibrium and non-equilibrium structures of small organic\nmolecules with up to seven non-hydrogen (C, N, O, S, Cl) atoms. To span this\nfundamentally important region of chemical compound space (CCS), QM7-X includes\nan exhaustive sampling of (meta-)stable equilibrium structures - comprised of\nconstitutional/structural isomers and stereoisomers, e.g., enantiomers and\ndiastereomers (including cis-/trans- and conformational isomers) - as well as\n100 non-equilibrium structural variations thereof to reach a total of $\\approx$\n4.2 M molecular structures. Computed at the tightly converged\nquantum-mechanical PBE0+MBD level of theory, QM7-X contains global (molecular)\nand local (atom-in-a-molecule) properties ranging from ground state quantities\n(such as atomization energies and dipole moments) to response quantities (such\nas polarizability tensors and dispersion coefficients). By providing a\nsystematic, extensive, and tightly-converged dataset of quantum-mechanically\ncomputed physicochemical properties, we expect that QM7-X will play a critical\nrole in the development of next-generation machine-learning based models for\nexploring greater swaths of CCS and performing in silico design of molecules\nwith targeted properties.", "title": "QM7-X: A comprehensive dataset of quantum-mechanical properties spanning the chemical space of small organic molecules", "url": "http://arxiv.org/abs/2006.15139v1" }

{ "text_length": 1483 }

{ "abstract": "Seeking health-related advice on the internet has become a common practice in\nthe digital era. Determining the trustworthiness of medical claims found online\nand finding appropriate evidence for this information is increasingly\nchallenging. Fact-checking has emerged as an approach to assess the veracity of\nfactual claims using evidence from credible knowledge sources. To help advance\nthe automation of this task, in this paper, we introduce a novel dataset of 750\nhealth-related claims, labeled for veracity by medical experts and backed with\nevidence from appropriate clinical studies. We provide an analysis of the\ndataset, highlighting its characteristics and challenges. The dataset can be\nused for Machine Learning tasks related to automated fact-checking such as\nevidence retrieval, veracity prediction, and explanation generation. For this\npurpose, we provide baseline models based on different approaches, examine\ntheir performance, and discuss the findings.", "title": "HealthFC: A Dataset of Health Claims for Evidence-Based Medical Fact-Checking", "url": "http://arxiv.org/abs/2309.08503v1" }

{ "text_length": 1081 }

{ "abstract": "Sentiment analysis aims to extract people's emotions and opinion from their\ncomments on the web. It widely used in businesses to detect sentiment in social\ndata, gauge brand reputation, and understand customers. Most of articles in\nthis area have concentrated on the English language whereas there are limited\nresources for Persian language. In this review paper, recent published articles\nbetween 2018 and 2022 in sentiment analysis in Persian Language have been\ncollected and their methods, approach and dataset will be explained and\nanalyzed. Almost all the methods used to solve sentiment analysis are machine\nlearning and deep learning. The purpose of this paper is to examine 40\ndifferent approach sentiment analysis in the Persian Language, analysis\ndatasets along with the accuracy of the algorithms applied to them and also\nreview strengths and weaknesses of each. Among all the methods, transformers\nsuch as BERT and RNN Neural Networks such as LSTM and Bi-LSTM have achieved\nhigher accuracy in the sentiment analysis. In addition to the methods and\napproaches, the datasets reviewed are listed between 2018 and 2022 and\ninformation about each dataset and its details are provided.", "title": "Sentiment Analysis of Persian Language: Review of Algorithms, Approaches and Datasets", "url": "http://arxiv.org/abs/2212.06041v1" }

{ "text_length": 1311 }

{ "abstract": "Crime is an unlawful act that carries legal repercussions. Bangladesh has a\nhigh crime rate due to poverty, population growth, and many other\nsocio-economic issues. For law enforcement agencies, understanding crime\npatterns is essential for preventing future criminal activity. For this\npurpose, these agencies need structured crime database. This paper introduces a\nnovel crime dataset that contains temporal, geographic, weather, and\ndemographic data about 6574 crime incidents of Bangladesh. We manually gather\ncrime news articles of a seven year time span from a daily newspaper archive.\nWe extract basic features from these raw text. Using these basic features, we\nthen consult standard service-providers of geo-location and weather data in\norder to garner these information related to the collected crime incidents.\nFurthermore, we collect demographic information from Bangladesh National Census\ndata. All these information are combined that results in a standard machine\nlearning dataset. Together, 36 features are engineered for the crime prediction\ntask. Five supervised machine learning classification algorithms are then\nevaluated on this newly built dataset and satisfactory results are achieved. We\nalso conduct exploratory analysis on various aspects the dataset. This dataset\nis expected to serve as the foundation for crime incidence prediction systems\nfor Bangladesh and other countries. The findings of this study will help law\nenforcement agencies to forecast and contain crime as well as to ensure optimal\nresource allocation for crime patrol and prevention.", "title": "Crime Prediction using Machine Learning with a Novel Crime Dataset", "url": "http://arxiv.org/abs/2211.01551v1" }

{ "text_length": 1679 }

{ "abstract": "To assist in the development of machine learning methods for automated\nclassification of spectroscopic data, we have generated a universal synthetic\ndataset that can be used for model validation. This dataset contains artificial\nspectra designed to represent experimental measurements from techniques\nincluding X-ray diffraction, nuclear magnetic resonance, and Raman\nspectroscopy. The dataset generation process features customizable parameters,\nsuch as scan length and peak count, which can be adjusted to fit the problem at\nhand. As an initial benchmark, we simulated a dataset containing 35,000 spectra\nbased on 500 unique classes. To automate the classification of this data, eight\ndifferent machine learning architectures were evaluated. From the results, we\nshed light on which factors are most critical to achieve optimal performance\nfor the classification task. The scripts used to generate synthetic spectra, as\nwell as our benchmark dataset and evaluation routines, are made publicly\navailable to aid in the development of improved machine learning models for\nspectroscopic analysis.", "title": "A universal synthetic dataset for machine learning on spectroscopic data", "url": "http://arxiv.org/abs/2206.06031v2" }

{ "text_length": 1201 }

{ "abstract": "Generalizability is the ultimate goal of Machine Learning (ML) image\nclassifiers, for which noise and limited dataset size are among the major\nconcerns. We tackle these challenges through utilizing the framework of deep\nMultitask Learning (dMTL) and incorporating image depth estimation as an\nauxiliary task. On a customized and depth-augmented derivation of the MNIST\ndataset, we show a) multitask loss functions are the most effective approach of\nimplementing dMTL, b) limited dataset size primarily contributes to\nclassification inaccuracy, and c) depth estimation is mostly impacted by noise.\nIn order to further validate the results, we manually labeled the NYU Depth V2\ndataset for scene classification tasks. As a contribution to the field, we have\nmade the data in python native format publicly available as an open-source\ndataset and provided the scene labels. Our experiments on MNIST and\nNYU-Depth-V2 show dMTL improves generalizability of the classifiers when the\ndataset is noisy and the number of examples is limited.", "title": "Minimizing the Effect of Noise and Limited Dataset Size in Image Classification Using Depth Estimation as an Auxiliary Task with Deep Multitask Learning", "url": "http://arxiv.org/abs/2208.10390v1" }

{ "text_length": 1218 }

{ "abstract": "Video-quality measurement is a critical task in video processing. Nowadays,\nmany implementations of new encoding standards - such as AV1, VVC, and LCEVC -\nuse deep-learning-based decoding algorithms with perceptual metrics that serve\nas optimization objectives. But investigations of the performance of modern\nvideo- and image-quality metrics commonly employ videos compressed using older\nstandards, such as AVC. In this paper, we present a new benchmark for\nvideo-quality metrics that evaluates video compression. It is based on a new\ndataset consisting of about 2,500 streams encoded using different standards,\nincluding AVC, HEVC, AV1, VP9, and VVC. Subjective scores were collected using\ncrowdsourced pairwise comparisons. The list of evaluated metrics includes\nrecent ones based on machine learning and neural networks. The results\ndemonstrate that new no-reference metrics exhibit a high correlation with\nsubjective quality and approach the capability of top full-reference metrics.", "title": "Video compression dataset and benchmark of learning-based video-quality metrics", "url": "http://arxiv.org/abs/2211.12109v2" }

{ "text_length": 1102 }

{ "abstract": "Falls are one of the leading cause of injury-related deaths among the elderly\nworldwide. Effective detection of falls can reduce the risk of complications\nand injuries. Fall detection can be performed using wearable devices or ambient\nsensors; these methods may struggle with user compliance issues or false\nalarms. Video cameras provide a passive alternative; however, regular RGB\ncameras are impacted by changing lighting conditions and privacy concerns. From\na machine learning perspective, developing an effective fall detection system\nis challenging because of the rarity and variability of falls. Many existing\nfall detection datasets lack important real-world considerations, such as\nvaried lighting, continuous activities of daily living (ADLs), and camera\nplacement. The lack of these considerations makes it difficult to develop\npredictive models that can operate effectively in the real world. To address\nthese limitations, we introduce a novel multi-modality dataset (MUVIM) that\ncontains four visual modalities: infra-red, depth, RGB and thermal cameras.\nThese modalities offer benefits such as obfuscated facial features and improved\nperformance in low-light conditions. We formulated fall detection as an anomaly\ndetection problem, in which a customized spatio-temporal convolutional\nautoencoder was trained only on ADLs so that a fall would increase the\nreconstruction error. Our results showed that infra-red cameras provided the\nhighest level of performance (AUC ROC=0.94), followed by thermal (AUC\nROC=0.87), depth (AUC ROC=0.86) and RGB (AUC ROC=0.83). This research provides\na unique opportunity to analyze the utility of camera modalities in detecting\nfalls in a home setting while balancing performance, passiveness, and privacy.", "title": "Multi Visual Modality Fall Detection Dataset", "url": "http://arxiv.org/abs/2206.12740v1" }

{ "text_length": 1831 }

{ "abstract": "Social media platforms and online streaming services have spawned a new breed\nof Hate Speech (HS). Due to the massive amount of user-generated content on\nthese sites, modern machine learning techniques are found to be feasible and\ncost-effective to tackle this problem. However, linguistically diverse datasets\ncovering different social contexts in which offensive language is typically\nused are required to train generalizable models. In this paper, we identify the\nshortcomings of existing Bangla HS datasets and introduce a large manually\nlabeled dataset BD-SHS that includes HS in different social contexts. The\nlabeling criteria were prepared following a hierarchical annotation process,\nwhich is the first of its kind in Bangla HS to the best of our knowledge. The\ndataset includes more than 50,200 offensive comments crawled from online social\nnetworking sites and is at least 60% larger than any existing Bangla HS\ndatasets. We present the benchmark result of our dataset by training different\nNLP models resulting in the best one achieving an F1-score of 91.0%. In our\nexperiments, we found that a word embedding trained exclusively using 1.47\nmillion comments from social media and streaming sites consistently resulted in\nbetter modeling of HS detection in comparison to other pre-trained embeddings.\nOur dataset and all accompanying codes is publicly available at\ngithub.com/naurosromim/hate-speech-dataset-for-Bengali-social-media", "title": "BD-SHS: A Benchmark Dataset for Learning to Detect Online Bangla Hate Speech in Different Social Contexts", "url": "http://arxiv.org/abs/2206.00372v1" }

{ "text_length": 1583 }

{ "abstract": "Semantic Machines (SM) have introduced the use of the dataflow (DF) paradigm\nto dialogue modelling, using computational graphs to hierarchically represent\nuser requests, data, and the dialogue history [Semantic Machines et al. 2020].\nAlthough the main focus of that paper was the SMCalFlow dataset (to date, the\nonly dataset with \"native\" DF annotations), they also reported some results of\nan experiment using a transformed version of the commonly used MultiWOZ dataset\n[Budzianowski et al. 2018] into a DF format. In this paper, we expand the\nexperiments using DF for the MultiWOZ dataset, exploring some additional\nexperimental set-ups. The code and instructions to reproduce the experiments\nreported here have been released. The contributions of this paper are: 1.) A DF\nimplementation capable of executing MultiWOZ dialogues; 2.) Several versions of\nconversion of MultiWOZ into a DF format are presented; 3.) Experimental results\non state match and translation accuracy.", "title": "MultiWOZ-DF -- A Dataflow implementation of the MultiWOZ dataset", "url": "http://arxiv.org/abs/2211.02303v1" }

{ "text_length": 1074 }

{ "abstract": "In electronic trading markets, limit order books (LOBs) provide information\nabout pending buy/sell orders at various price levels for a given security.\nRecently, there has been a growing interest in using LOB data for resolving\ndownstream machine learning tasks (e.g., forecasting). However, dealing with\nout-of-distribution (OOD) LOB data is challenging since distributional shifts\nare unlabeled in current publicly available LOB datasets. Therefore, it is\ncritical to build a synthetic LOB dataset with labeled OOD samples serving as a\ntestbed for developing models that generalize well to unseen scenarios. In this\nwork, we utilize a multi-agent market simulator to build a synthetic LOB\ndataset, named DSLOB, with and without market stress scenarios, which allows\nfor the design of controlled distributional shift benchmarking. Using the\nproposed synthetic dataset, we provide a holistic analysis on the forecasting\nperformance of three different state-of-the-art forecasting methods. Our\nresults reflect the need for increased researcher efforts to develop algorithms\nwith robustness to distributional shifts in high-frequency time series data.", "title": "DSLOB: A Synthetic Limit Order Book Dataset for Benchmarking Forecasting Algorithms under Distributional Shift", "url": "http://arxiv.org/abs/2211.11513v1" }

{ "text_length": 1294 }

{ "abstract": "As restricted quantum computers are slowly becoming a reality, the search for\nmeaningful first applications intensifies. In this domain, one of the more\ninvestigated approaches is the use of a special type of quantum circuit - a\nso-called quantum neural network -- to serve as a basis for a machine learning\nmodel. Roughly speaking, as the name suggests, a quantum neural network can\nplay a similar role to a neural network. However, specifically for applications\nin machine learning contexts, very little is known about suitable circuit\narchitectures, or model hyperparameters one should use to achieve good learning\nperformance. In this work, we apply the functional ANOVA framework to quantum\nneural networks to analyze which of the hyperparameters were most influential\nfor their predictive performance. We analyze one of the most typically used\nquantum neural network architectures. We then apply this to $7$ open-source\ndatasets from the OpenML-CC18 classification benchmark whose number of features\nis small enough to fit on quantum hardware with less than $20$ qubits. Three\nmain levels of importance were detected from the ranking of hyperparameters\nobtained with functional ANOVA. Our experiment both confirmed expected patterns\nand revealed new insights. For instance, setting well the learning rate is\ndeemed the most critical hyperparameter in terms of marginal contribution on\nall datasets, whereas the particular choice of entangling gates used is\nconsidered the least important except on one dataset. This work introduces new\nmethodologies to study quantum machine learning models and provides new\ninsights toward quantum model selection.", "title": "Hyperparameter Importance of Quantum Neural Networks Across Small Datasets", "url": "http://arxiv.org/abs/2206.09992v1" }

{ "text_length": 1763 }

{ "abstract": "Social media is often utilized as a lifeline for communication during natural\ndisasters. Traditionally, natural disaster tweets are filtered from the Twitter\nstream using the name of the natural disaster and the filtered tweets are sent\nfor human annotation. The process of human annotation to create labeled sets\nfor machine learning models is laborious, time consuming, at times inaccurate,\nand more importantly not scalable in terms of size and real-time use. In this\nwork, we curate a silver standard dataset using weak supervision. In order to\nvalidate its utility, we train machine learning models on the weakly supervised\ndata to identify three different types of natural disasters i.e earthquakes,\nhurricanes and floods. Our results demonstrate that models trained on the\nsilver standard dataset achieved performance greater than 90% when classifying\na manually curated, gold-standard dataset. To enable reproducible research and\nadditional downstream utility, we release the silver standard dataset for the\nscientific community.", "title": "TweetDIS: A Large Twitter Dataset for Natural Disasters Built using Weak Supervision", "url": "http://arxiv.org/abs/2207.04947v1" }

{ "text_length": 1156 }

{ "abstract": "We introduce MANTRA, an annotated dataset of 4869 transient and 71207\nnon-transient object lightcurves built from the Catalina Real Time Transient\nSurvey. We provide public access to this dataset as a plain text file to\nfacilitate standardized quantitative comparison of astronomical transient event\nrecognition algorithms. Some of the classes included in the dataset are:\nsupernovae, cataclysmic variables, active galactic nuclei, high proper motion\nstars, blazars and flares. As an example of the tasks that can be performed on\nthe dataset we experiment with multiple data pre-processing methods, feature\nselection techniques and popular machine learning algorithms (Support Vector\nMachines, Random Forests and Neural Networks). We assess quantitative\nperformance in two classification tasks: binary (transient/non-transient) and\neight-class classification. The best performing algorithm in both tasks is the\nRandom Forest Classifier. It achieves an F1-score of 96.25% in the binary\nclassification and 52.79% in the eight-class classification. For the\neight-class classification, non-transients ( 96.83% ) is the class with the\nhighest F1-score, while the lowest corresponds to high-proper-motion stars (\n16.79% ); for supernovae it achieves a value of 54.57% , close to the average\nacross classes. The next release of MANTRA includes images and benchmarks with\ndeep learning models.", "title": "MANTRA: A Machine Learning reference lightcurve dataset for astronomical transient event recognition", "url": "http://arxiv.org/abs/2006.13163v2" }

{ "text_length": 1520 }

{ "abstract": "We introduce ArtBench-10, the first class-balanced, high-quality, cleanly\nannotated, and standardized dataset for benchmarking artwork generation. It\ncomprises 60,000 images of artwork from 10 distinctive artistic styles, with\n5,000 training images and 1,000 testing images per style. ArtBench-10 has\nseveral advantages over previous artwork datasets. Firstly, it is\nclass-balanced while most previous artwork datasets suffer from the long tail\nclass distributions. Secondly, the images are of high quality with clean\nannotations. Thirdly, ArtBench-10 is created with standardized data collection,\nannotation, filtering, and preprocessing procedures. We provide three versions\nof the dataset with different resolutions ($32\\times32$, $256\\times256$, and\noriginal image size), formatted in a way that is easy to be incorporated by\npopular machine learning frameworks. We also conduct extensive benchmarking\nexperiments using representative image synthesis models with ArtBench-10 and\npresent in-depth analysis. The dataset is available at\nhttps://github.com/liaopeiyuan/artbench under a Fair Use license.", "title": "The ArtBench Dataset: Benchmarking Generative Models with Artworks", "url": "http://arxiv.org/abs/2206.11404v1" }

{ "text_length": 1204 }

{ "abstract": "Unlike human daily activities, existing publicly available sensor datasets\nfor work activity recognition in industrial domains are limited by difficulties\nin collecting realistic data as close collaboration with industrial sites is\nrequired. This also limits research on and development of AI methods for\nindustrial applications. To address these challenges and contribute to research\non machine recognition of work activities in industrial domains, in this study,\nwe introduce a new large-scale dataset for packaging work recognition called\nOpenPack. OpenPack contains 53.8 hours of multimodal sensor data, including\nkeypoints, depth images, acceleration data, and readings from IoT-enabled\ndevices (e.g., handheld barcode scanners used in work procedures), collected\nfrom 16 distinct subjects with different levels of packaging work experience.\nOn the basis of this dataset, we propose a neural network model designed to\nrecognize work activities, which efficiently fuses sensor data and readings\nfrom IoT-enabled devices by processing them within different streams in a\nladder-shaped architecture, and the experiment showed the effectiveness of the\narchitecture. We believe that OpenPack will contribute to the community of\naction/activity recognition with sensors. OpenPack dataset is available at\nhttps://open-pack.github.io/.", "title": "OpenPack: A Large-scale Dataset for Recognizing Packaging Works in IoT-enabled Logistic Environments", "url": "http://arxiv.org/abs/2212.11152v1" }

{ "text_length": 1466 }

{ "abstract": "Allowing organizations to share their data for training of machine learning\n(ML) models without unintended information leakage is an open problem in\npractice. A promising technique for this still-open problem is to train models\non the encoded data. Our approach, called Privately Encoded Open Datasets with\nPublic Labels (PEOPL), uses a certain class of randomly constructed transforms\nto encode sensitive data. Organizations publish their randomly encoded data and\nassociated raw labels for ML training, where training is done without knowledge\nof the encoding realization. We investigate several important aspects of this\nproblem: We introduce information-theoretic scores for privacy and utility,\nwhich quantify the average performance of an unfaithful user (e.g., adversary)\nand a faithful user (e.g., model developer) that have access to the published\nencoded data. We then theoretically characterize primitives in building\nfamilies of encoding schemes that motivate the use of random deep neural\nnetworks. Empirically, we compare the performance of our randomized encoding\nscheme and a linear scheme to a suite of computational attacks, and we also\nshow that our scheme achieves competitive prediction accuracy to raw-sample\nbaselines. Moreover, we demonstrate that multiple institutions, using\nindependent random encoders, can collaborate to train improved ML models.", "title": "PEOPL: Characterizing Privately Encoded Open Datasets with Public Labels", "url": "http://arxiv.org/abs/2304.00047v1" }

{ "text_length": 1481 }

{ "abstract": "Predicting the probability of non-performing loans for individuals has a\nvital and beneficial role for banks to decrease credit risk and make the right\ndecisions before giving the loan. The trend to make these decisions are based\non credit study and in accordance with generally accepted standards, loan\npayment history, and demographic data of the clients. In this work, we evaluate\nhow different Machine learning models such as Random Forest, Decision tree,\nKNN, SVM, and XGBoost perform on the dataset provided by a private bank in\nEthiopia. Further, motivated by this evaluation we explore different feature\nselection methods to state the important features for the bank. Our findings\nshow that XGBoost achieves the highest F1 score on the KMeans SMOTE\nover-sampled data. We also found that the most important features are the age\nof the applicant, years of employment, and total income of the applicant rather\nthan collateral-related features in evaluating credit risk.", "title": "Machine Learning Models Evaluation and Feature Importance Analysis on NPL Dataset", "url": "http://arxiv.org/abs/2209.09638v1" }

{ "text_length": 1090 }

{ "abstract": "As free online encyclopedias with massive volumes of content, Wikipedia and\nWikidata are key to many Natural Language Processing (NLP) tasks, such as\ninformation retrieval, knowledge base building, machine translation, text\nclassification, and text summarization. In this paper, we introduce WikiDes, a\nnovel dataset to generate short descriptions of Wikipedia articles for the\nproblem of text summarization. The dataset consists of over 80k English samples\non 6987 topics. We set up a two-phase summarization method - description\ngeneration (Phase I) and candidate ranking (Phase II) - as a strong approach\nthat relies on transfer and contrastive learning. For description generation,\nT5 and BART show their superiority compared to other small-scale pre-trained\nmodels. By applying contrastive learning with the diverse input from beam\nsearch, the metric fusion-based ranking models outperform the direct\ndescription generation models significantly up to 22 ROUGE in topic-exclusive\nsplit and topic-independent split. Furthermore, the outcome descriptions in\nPhase II are supported by human evaluation in over 45.33% chosen compared to\n23.66% in Phase I against the gold descriptions. In the aspect of sentiment\nanalysis, the generated descriptions cannot effectively capture all sentiment\npolarities from paragraphs while doing this task better from the gold\ndescriptions. The automatic generation of new descriptions reduces the human\nefforts in creating them and enriches Wikidata-based knowledge graphs. Our\npaper shows a practical impact on Wikipedia and Wikidata since there are\nthousands of missing descriptions. Finally, we expect WikiDes to be a useful\ndataset for related works in capturing salient information from short\nparagraphs. The curated dataset is publicly available at:\nhttps://github.com/declare-lab/WikiDes.", "title": "WikiDes: A Wikipedia-Based Dataset for Generating Short Descriptions from Paragraphs", "url": "http://arxiv.org/abs/2209.13101v1" }

{ "text_length": 1949 }

{ "abstract": "We establish a dataset of over $1.6\\times10^4$ experimental images of\nBose--Einstein condensates containing solitonic excitations to enable machine\nlearning (ML) for many-body physics research. About $33~\\%$ of this dataset has\nmanually assigned and carefully curated labels. The remainder is automatically\nlabeled using SolDet -- an implementation of a physics-informed ML data\nanalysis framework -- consisting of a convolutional-neural-network-based\nclassifier and OD as well as a statistically motivated physics-informed\nclassifier and a quality metric. This technical note constitutes the definitive\nreference of the dataset, providing an opportunity for the data science\ncommunity to develop more sophisticated analysis tools, to further understand\nnonlinear many-body physics, and even advance cold atom experiments.", "title": "Dark solitons in Bose-Einstein condensates: a dataset for many-body physics research", "url": "http://arxiv.org/abs/2205.09114v2" }

{ "text_length": 941 }

{ "abstract": "Subseasonal forecasting of the weather two to six weeks in advance is\ncritical for resource allocation and climate adaptation but poses many\nchallenges for the forecasting community. At this forecast horizon,\nphysics-based dynamical models have limited skill, and the targets for\nprediction depend in a complex manner on both local weather and global climate\nvariables. Recently, machine learning methods have shown promise in advancing\nthe state of the art but only at the cost of complex data curation, integrating\nexpert knowledge with aggregation across multiple relevant data sources, file\nformats, and temporal and spatial resolutions. To streamline this process and\naccelerate future development, we introduce SubseasonalClimateUSA, a curated\ndataset for training and benchmarking subseasonal forecasting models in the\nUnited States. We use this dataset to benchmark a diverse suite of subseasonal\nmodels, including operational dynamical models, classical meteorological\nbaselines, and ten state-of-the-art machine learning and deep learning-based\nmethods from the literature. Overall, our benchmarks suggest simple and\neffective ways to extend the accuracy of current operational models.\nSubseasonalClimateUSA is regularly updated and accessible via the\nhttps://github.com/microsoft/subseasonal_data/ Python package.", "title": "SubseasonalClimateUSA: A Dataset for Subseasonal Forecasting and Benchmarking", "url": "http://arxiv.org/abs/2109.10399v3" }

{ "text_length": 1436 }

{ "abstract": "Clinical diagnosis of the eye is performed over multifarious data modalities\nincluding scalar clinical labels, vectorized biomarkers, two-dimensional fundus\nimages, and three-dimensional Optical Coherence Tomography (OCT) scans.\nClinical practitioners use all available data modalities for diagnosing and\ntreating eye diseases like Diabetic Retinopathy (DR) or Diabetic Macular Edema\n(DME). Enabling usage of machine learning algorithms within the ophthalmic\nmedical domain requires research into the relationships and interactions\nbetween all relevant data over a treatment period. Existing datasets are\nlimited in that they neither provide data nor consider the explicit\nrelationship modeling between the data modalities. In this paper, we introduce\nthe Ophthalmic Labels for Investigating Visual Eye Semantics (OLIVES) dataset\nthat addresses the above limitation. This is the first OCT and near-IR fundus\ndataset that includes clinical labels, biomarker labels, disease labels, and\ntime-series patient treatment information from associated clinical trials. The\ndataset consists of 1268 near-IR fundus images each with at least 49 OCT scans,\nand 16 biomarkers, along with 4 clinical labels and a disease diagnosis of DR\nor DME. In total, there are 96 eyes' data averaged over a period of at least\ntwo years with each eye treated for an average of 66 weeks and 7 injections. We\nbenchmark the utility of OLIVES dataset for ophthalmic data as well as provide\nbenchmarks and concrete research directions for core and emerging machine\nlearning paradigms within medical image analysis.", "title": "OLIVES Dataset: Ophthalmic Labels for Investigating Visual Eye Semantics", "url": "http://arxiv.org/abs/2209.11195v1" }

{ "text_length": 1688 }

{ "abstract": "The FAIR Guiding Principles aim to improve the findability, accessibility,\ninteroperability, and reusability of digital content by making them both human\nand machine actionable. However, these principles have not yet been broadly\nadopted in the domain of machine learning-based program analyses and\noptimizations for High-Performance Computing (HPC). In this paper, we design a\nmethodology to make HPC datasets and machine learning models FAIR after\ninvestigating existing FAIRness assessment and improvement techniques. Our\nmethodology includes a comprehensive, quantitative assessment for elected data,\nfollowed by concrete, actionable suggestions to improve FAIRness with respect\nto common issues related to persistent identifiers, rich metadata descriptions,\nlicense and provenance information. Moreover, we select a representative\ntraining dataset to evaluate our methodology. The experiment shows the\nmethodology can effectively improve the dataset and model's FAIRness from an\ninitial score of 19.1% to the final score of 83.0%.", "title": "Making Machine Learning Datasets and Models FAIR for HPC: A Methodology and Case Study", "url": "http://arxiv.org/abs/2211.02092v1" }

{ "text_length": 1156 }

{ "abstract": "Agriculture is of one of the few remaining sectors that is yet to receive\nproper attention from the machine learning community. The importance of\ndatasets in the machine learning discipline cannot be overemphasized. The lack\nof standard and publicly available datasets related to agriculture impedes\npractitioners of this discipline to harness the full benefit of these powerful\ncomputational predictive tools and techniques. To improve this scenario, we\ndevelop, to the best of our knowledge, the first-ever standard, ready-to-use,\nand publicly available dataset of mango leaves. The images are collected from\nfour mango orchards of Bangladesh, one of the top mango-growing countries of\nthe world. The dataset contains 4000 images of about 1800 distinct leaves\ncovering seven diseases. Although the dataset is developed using mango leaves\nof Bangladesh only, since we deal with diseases that are common across many\ncountries, this dataset is likely to be applicable to identify mango diseases\nin other countries as well, thereby boosting mango yield. This dataset is\nexpected to draw wide attention from machine learning researchers and\npractitioners in the field of automated agriculture.", "title": "MangoLeafBD: A Comprehensive Image Dataset to Classify Diseased and Healthy Mango Leaves", "url": "http://arxiv.org/abs/2209.02377v1" }

{ "text_length": 1313 }

{ "abstract": "Machine reading comprehension (MRC) is a crucial task in natural language\nprocessing and has achieved remarkable advancements. However, most of the\nneural MRC models are still far from robust and fail to generalize well in\nreal-world applications. In order to comprehensively verify the robustness and\ngeneralization of MRC models, we introduce a real-world Chinese dataset --\nDuReader_robust. It is designed to evaluate the MRC models from three aspects:\nover-sensitivity, over-stability and generalization. Comparing to previous\nwork, the instances in DuReader_robust are natural texts, rather than the\naltered unnatural texts. It presents the challenges when applying MRC models to\nreal-world applications. The experimental results show that MRC models do not\nperform well on the challenge test set. Moreover, we analyze the behavior of\nexisting models on the challenge test set, which may provide suggestions for\nfuture model development. The dataset and codes are publicly available at\nhttps://github.com/baidu/DuReader.", "title": "DuReader_robust: A Chinese Dataset Towards Evaluating Robustness and Generalization of Machine Reading Comprehension in Real-World Applications", "url": "http://arxiv.org/abs/2004.11142v2" }

{ "text_length": 1203 }

{ "abstract": "An important task within the broader goal of Space Situational Awareness\n(SSA) is to observe changes in the orbits of satellites, where the data spans\nthousands of objects over long time scales (decades). The Two-Line Element\n(TLE) data provided by the North American Aerospace Defense Command is the most\ncomprehensive and widely-available dataset cataloguing the orbits of\nsatellites. This makes it a highly-attractive data source on which to perform\nthis observation. However, when attempting to infer changes in satellite\nbehaviour from TLE data, there are a number of potential pitfalls. These mostly\nrelate to specific features of the TLE data which are not always clearly\ndocumented in the data sources or popular software packages for manipulating\nthem. These quirks produce a particularly hazardous data type for researchers\nfrom adjacent disciplines (such as anomaly detection or machine learning). We\nhighlight these features of TLE data and the resulting pitfalls in order to\nsave future researchers from being trapped. A seperate, significant, issue is\nthat existing contributions to manoeuvre detection from TLE data evaluate their\nalgorithms on different satellites, making comparison between these methods\ndifficult. Moreover, the ground-truth in these datasets is often poor quality,\nsometimes being based on subjective human assessment. We therefore release and\ndescribe in-depth an open, curated, benchmark dataset containing TLE data for\n15 satellites alongside high-quality ground-truth manoeuvre timestamps.", "title": "Wide-scale Monitoring of Satellite Lifetimes: Pitfalls and a Benchmark Dataset", "url": "http://arxiv.org/abs/2212.08662v1" }

{ "text_length": 1642 }

{ "abstract": "A riddle is a question or statement with double or veiled meanings, followed\nby an unexpected answer. Solving riddle is a challenging task for both machine\nand human, testing the capability of understanding figurative, creative natural\nlanguage and reasoning with commonsense knowledge. We introduce BiRdQA, a\nbilingual multiple-choice question answering dataset with 6614 English riddles\nand 8751 Chinese riddles. For each riddle-answer pair, we provide four\ndistractors with additional information from Wikipedia. The distractors are\nautomatically generated at scale with minimal bias. Existing monolingual and\nmultilingual QA models fail to perform well on our dataset, indicating that\nthere is a long way to go before machine can beat human on solving tricky\nriddles. The dataset has been released to the community.", "title": "BiRdQA: A Bilingual Dataset for Question Answering on Tricky Riddles", "url": "http://arxiv.org/abs/2109.11087v2" }

{ "text_length": 922 }

{ "abstract": "Smart manufacturing systems are being deployed at a growing rate because of\ntheir ability to interpret a wide variety of sensed information and act on the\nknowledge gleaned from system observations. In many cases, the principal goal\nof the smart manufacturing system is to rapidly detect (or anticipate) failures\nto reduce operational cost and eliminate downtime. This often boils down to\ndetecting anomalies within the sensor date acquired from the system. The smart\nmanufacturing application domain poses certain salient technical challenges. In\nparticular, there are often multiple types of sensors with varying capabilities\nand costs. The sensor data characteristics change with the operating point of\nthe environment or machines, such as, the RPM of the motor. The anomaly\ndetection process therefore has to be calibrated near an operating point. In\nthis paper, we analyze four datasets from sensors deployed from manufacturing\ntestbeds. We evaluate the performance of several traditional and ML-based\nforecasting models for predicting the time series of sensor data. Then,\nconsidering the sparse data from one kind of sensor, we perform transfer\nlearning from a high data rate sensor to perform defect type classification.\nTaken together, we show that predictive failure classification can be achieved,\nthus paving the way for predictive maintenance.", "title": "Anomaly Detection and Inter-Sensor Transfer Learning on Smart Manufacturing Datasets", "url": "http://arxiv.org/abs/2206.06355v1" }

{ "text_length": 1475 }

{ "abstract": "Lecture slide presentations, a sequence of pages that contain text and\nfigures accompanied by speech, are constructed and presented carefully in order\nto optimally transfer knowledge to students. Previous studies in multimedia and\npsychology attribute the effectiveness of lecture presentations to their\nmultimodal nature. As a step toward developing AI to aid in student learning as\nintelligent teacher assistants, we introduce the Multimodal Lecture\nPresentations dataset as a large-scale benchmark testing the capabilities of\nmachine learning models in multimodal understanding of educational content. Our\ndataset contains aligned slides and spoken language, for 180+ hours of video\nand 9000+ slides, with 10 lecturers from various subjects (e.g., computer\nscience, dentistry, biology). We introduce two research tasks which are\ndesigned as stepping stones towards AI agents that can explain (automatically\ncaptioning a lecture presentation) and illustrate (synthesizing visual figures\nto accompany spoken explanations) educational content. We provide manual\nannotations to help implement these two research tasks and evaluate\nstate-of-the-art models on them. Comparing baselines and human student\nperformances, we find that current models struggle in (1) weak crossmodal\nalignment between slides and spoken text, (2) learning novel visual mediums,\n(3) technical language, and (4) long-range sequences. Towards addressing this\nissue, we also introduce PolyViLT, a multimodal transformer trained with a\nmulti-instance learning loss that is more effective than current approaches. We\nconclude by shedding light on the challenges and opportunities in multimodal\nunderstanding of educational presentations.", "title": "Multimodal Lecture Presentations Dataset: Understanding Multimodality in Educational Slides", "url": "http://arxiv.org/abs/2208.08080v1" }

{ "text_length": 1831 }

{ "abstract": "Machine learning (ML) models are costly to train as they can require a\nsignificant amount of data, computational resources and technical expertise.\nThus, they constitute valuable intellectual property that needs protection from\nadversaries wanting to steal them. Ownership verification techniques allow the\nvictims of model stealing attacks to demonstrate that a suspect model was in\nfact stolen from theirs.\n Although a number of ownership verification techniques based on watermarking\nor fingerprinting have been proposed, most of them fall short either in terms\nof security guarantees (well-equipped adversaries can evade verification) or\ncomputational cost. A fingerprinting technique, Dataset Inference (DI), has\nbeen shown to offer better robustness and efficiency than prior methods.\n The authors of DI provided a correctness proof for linear (suspect) models.\nHowever, in a subspace of the same setting, we prove that DI suffers from high\nfalse positives (FPs) -- it can incorrectly identify an independent model\ntrained with non-overlapping data from the same distribution as stolen. We\nfurther prove that DI also triggers FPs in realistic, non-linear suspect\nmodels. We then confirm empirically that DI in the black-box setting leads to\nFPs, with high confidence.\n Second, we show that DI also suffers from false negatives (FNs) -- an\nadversary can fool DI (at the cost of incurring some accuracy loss) by\nregularising a stolen model's decision boundaries using adversarial training,\nthereby leading to an FN. To this end, we demonstrate that black-box DI fails\nto identify a model adversarially trained from a stolen dataset -- the setting\nwhere DI is the hardest to evade.\n Finally, we discuss the implications of our findings, the viability of\nfingerprinting-based ownership verification in general, and suggest directions\nfor future work.", "title": "On the Robustness of Dataset Inference", "url": "http://arxiv.org/abs/2210.13631v3" }

{ "text_length": 1929 }