A common challenge across all areas of machine learning is that training data is not distributed like test data, due to natural shifts, "blind spots," or adversarial examples; such test examples are referred to as out-of-distribution (OOD) test examples. We consider a model where one may abstain from predicting, at a fixed cost. In particular, our transductive abstention algorithm takes labeled training examples and unlabeled test examples as input, and provides predictions with optimal prediction loss guarantees. The loss bounds match standard generalization bounds when test examples are i.i.d. from the training distribution, but add an additional term that is the cost of abstaining times the statistical distance between the train and test distribution (or the fraction of adversarial examples). For linear regression, we give a polynomial-time algorithm based on Celis-Dennis-Tapia optimization algorithms. For binary classification, we show how to efficiently implement it using a proper agnostic learner (i.e., an Empirical Risk Minimizer) for the class of interest. Our work builds on a recent abstention algorithm of Goldwasser, Kalais, and Montasser (2020) for transductive binary classification.
With the increased interest in machine learning and big data problems, the need for large amounts of labelled data has also grown. However, it is often infeasible to get experts to label all of this data, which leads many practitioners to crowdsourcing solutions. In this paper, we present new techniques to improve the quality of the labels while attempting to reduce the cost. The naive approach to assigning labels is to adopt a majority vote method, however, in the context of data labelling, this is not always ideal as data labellers are not equally reliable. One might, instead, give higher priority to certain labellers through some kind of weighted vote based on past performance. This paper investigates the use of more sophisticated methods, such as Bayesian inference, to measure the performance of the labellers as well as the confidence of each label. The methods we propose follow an iterative improvement algorithm which attempts to use the least amount of workers necessary to achieve the desired confidence in the inferred label. This paper explores simulated binary classification problems with simulated workers and questions to test the proposed methods. Our methods outperform the standard voting methods in both cost and accuracy while maintaining higher reliability when there is disagreement within the crowd.
Social network platforms are generally used to share positive, constructive, and insightful content. However, in recent times, people often get exposed to objectionable content like threat, identity attacks, hate speech, insults, obscene texts, offensive remarks or bullying. Existing work on toxic speech detection focuses on binary classification or on differentiating toxic speech among a small set of categories. This paper describes the system proposed by team Cisco for SemEval-2021 Task 5: Toxic Spans Detection, the first shared task focusing on detecting the spans in the text that attribute to its toxicity, in English language. We approach this problem primarily in two ways: a sequence tagging approach and a dependency parsing approach. In our sequence tagging approach we tag each token in a sentence under a particular tagging scheme. Our best performing architecture in this approach also proved to be our best performing architecture overall with an F1 score of 0.6922, thereby placing us 7th on the final evaluation phase leaderboard. We also explore a dependency parsing approach where we extract spans from the input sentence under the supervision of target span boundaries and rank our spans using a biaffine model. Finally, we also provide a detailed analysis of our results and model performance in our paper.
Recent work has shown that the training of a one-hidden-layer, scalar-output fully-connected ReLU neural network can be reformulated as a finite-dimensional convex program. Unfortunately, the scale of such a convex program grows exponentially in data size. In this work, we prove that a stochastic procedure with a linear complexity well approximates the exact formulation. Moreover, we derive a convex optimization approach to efficiently solve the "adversarial training" problem, which trains neural networks that are robust to adversarial input perturbations. Our method can be applied to binary classification and regression, and provides an alternative to the current adversarial training methods, such as Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD). We demonstrate in experiments that the proposed method achieves a noticeably better adversarial robustness and performance than the existing methods.
Interactive robots navigating photo-realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision-and-language navigation (VLN). In this paper, we present VISITRON, a multi-modal Transformer-based navigator better suited to the interactive regime inherent to Cooperative Vision-and-Dialog Navigation (CVDN). VISITRON is trained to: i) identify and associate object-level concepts and semantics between the environment and dialogue history, ii) identify when to interact vs. navigate via imitation learning of a binary classification head. We perform extensive pre-training and fine-tuning ablations with VISITRON to gain empirical insights and improve performance on CVDN. VISITRON's ability to identify when to interact leads to a natural generalization of the game-play mode introduced by Roman et al. (arXiv:2005.00728) for enabling the use of such models in different environments. VISITRON is competitive with models on the static CVDN leaderboard and attains state-of-the-art performance on the Success weighted by Path Length (SPL) metric.
The purpose of this research report is to present the our learning curve and the exposure to the Machine Learning life cycle, with the use of a Kaggle binary classification data set and taking to explore various techniques from pre-processing to the final optimization and model evaluation, also we highlight on the data imbalance issue and we discuss the different methods of handling that imbalance on the data level by over-sampling and under sampling not only to reach a balanced class representation but to improve the overall performance. This work also opens some gaps for future work.
The Neural Tangent Kernel (NTK) is the wide-network limit of a kernel defined using neural networks at initialization, whose embedding is the gradient of the output of the network with respect to its parameters. We study the "after kernel", which is defined using the same embedding, except after training, for neural networks with standard architectures, on binary classification problems extracted from MNIST and CIFAR-10, trained using SGD in a standard way. For some dataset-architecture pairs, after a few epochs of neural network training, a hard-margin SVM using the network's after kernel is much more accurate than when the network's initial kernel is used. For networks with an architecture similar to VGG, the after kernel is more "global", in the sense that it is less invariant to transformations of input images that disrupt the global structure of the image while leaving the local statistics largely intact. For fully connected networks, the after kernel is less global in this sense. The after kernel tends to be more invariant to small shifts, rotations and zooms; data augmentation does not improve these invariances. The (finite approximation to the) conjugate kernel, obtained using the last layer of hidden nodes, sometimes, but not always, provides a good approximation to the NTK and the after kernel. Training a network with a larger learning rate (while holding the training error constant) produces a better kernel, as measured by the test error of a hard-margin SVM. The after kernels of networks trained with larger learning rates tend to be more global, and more invariant to small shifts, rotations and zooms.
Artificial intelligence (AI) algorithms using deep learning have advanced the classification of skin disease images; however these algorithms have been mostly applied "in silico" and not validated clinically. Most dermatology AI algorithms perform binary classification tasks (e.g. malignancy versus benign lesions), but this task is not representative of dermatologists' diagnostic range. The American Academy of Dermatology Task Force on Augmented Intelligence published a position statement emphasizing the importance of clinical validation to create human-computer synergy, termed augmented intelligence (AuI). Liu et al's recent paper, "A deep learning system for differential diagnosis of skin diseases" represents a significant advancement of AI in dermatology, bringing it closer to clinical impact. However, significant issues must be addressed before this algorithm can be integrated into clinical workflow. These issues include accurate and equitable model development, defining and assessing appropriate clinical outcomes, and real-world integration.
Entanglement is a physical phenomenon, which has fueled recent successes of quantum algorithms. Although quantum neural networks (QNNs) have shown promising results in solving simple machine learning tasks recently, for the time being, the effect of entanglement in QNNs and the behavior of QNNs in binary pattern classification are still underexplored. In this work, we provide some theoretical insight into the properties of QNNs by presenting and analyzing a new form of invariance embedded in QNNs for both quantum binary classification and quantum representation learning, which we term negational symmetry. Given a quantum binary signal and its negational counterpart where a bitwise NOT operation is applied to each quantum bit of the binary signal, a QNN outputs the same logits. That is to say, QNNs cannot differentiate a quantum binary signal and its negational counterpart in a binary classification task. We further empirically evaluate the negational symmetry of QNNs in binary pattern classification tasks using Google's quantum computing framework. The theoretical and experimental results suggest that negational symmetry is a fundamental property of QNNs, which is not shared by classical models. Our findings also imply that negational symmetry is a double-edged sword in practical quantum applications.
Explainable AI has attracted much research attention in recent years with feature attribution algorithms, which compute "feature importance" in predictions, becoming increasingly popular. However, there is little analysis of the validity of these algorithms as there is no "ground truth" in the existing datasets to validate their correctness. In this work, we develop a method to quantitatively evaluate the correctness of XAI algorithms by creating datasets with known explanation ground truth. To this end, we focus on the binary classification problems. String datasets are constructed using formal language derived from a grammar. A string is positive if and only if a certain property is fulfilled. Symbols serving as explanation ground truth in a positive string are part of an explanation if and only if they contributes to fulfilling the property. Two popular feature attribution explainers, Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP), are used in our experiments.We show that: (1) classification accuracy is positively correlated with explanation accuracy; (2) SHAP provides more accurate explanations than LIME; (3) explanation accuracy is negatively correlated with dataset complexity.
In this paper, we study the problem of extracting variable-depth "logical document hierarchy" from long documents, namely organizing the recognized "physical document objects" into hierarchical structures. The discovery of logical document hierarchy is the vital step to support many downstream applications. However, long documents, containing hundreds or even thousands of pages and variable-depth hierarchy, challenge the existing methods. To address these challenges, we develop a framework, namely Hierarchy Extraction from Long Document (HELD), where we "sequentially" insert each physical object at the proper on of the current tree. Determining whether each possible position is proper or not can be formulated as a binary classification problem. To further improve its effectiveness and efficiency, we study the design variants in HELD, including traversal orders of the insertion positions, heading extraction explicitly or implicitly, tolerance to insertion errors in predecessor steps, and so on. The empirical experiments based on thousands of long documents from Chinese, English financial market and English scientific publication show that the HELD model with the "root-to-leaf" traversal order and explicit heading extraction is the best choice to achieve the tradeoff between effectiveness and efficiency with the accuracy of 0.9726, 0.7291 and 0.9578 in Chinese financial, English financial and arXiv datasets, respectively. Finally, we show that logical document hierarchy can be employed to significantly improve the performance of the downstream passage retrieval task. In summary, we conduct a systematic study on this task in terms of methods, evaluations, and applications.
Real-world binary classification tasks are in many cases imbalanced, where the minority class is much smaller than the majority class. This skewness is challenging for machine learning algorithms as they tend to focus on the majority and greatly misclassify the minority. Adding synthetic minority samples to the dataset before training the model is a popular technique to address this difficulty and is commonly achieved by interpolating minority samples. Tabular datasets are often multi-modal and contain discrete (categorical) features in addition to continuous ones which makes interpolation of samples non-trivial. To address this, we propose a latent space interpolation framework which (1) maps the multi-modal samples to a dense continuous latent space using an autoencoder; (2) applies oversampling by interpolation in the latent space; and (3) maps the synthetic samples back to the original feature space. We defined metrics to directly evaluate the quality of the minority data generated and showed that our framework generates better synthetic data than the existing methods. Furthermore, the superior synthetic data yields better prediction quality in downstream binary classification tasks, as was demonstrated in extensive experiments with 27 publicly available real-world datasets
Gene promoters are the key DNA regulatory elements positioned around the transcription start sites and are responsible for regulating gene transcription process. Various alignment-based, signal-based and content-based approaches are reported for the prediction of promoters. However, since all promoter sequences do not show explicit features, the prediction performance of these techniques is poor. Therefore, many machine learning and deep learning models have been proposed for promoter prediction. In this work, we studied methods for vector encoding and promoter classification using genome sequences of three distinct higher eukaryotes viz. yeast (Saccharomyces cerevisiae), A. thaliana (plant) and human (Homo sapiens). We compared one-hot vector encoding method with frequency-based tokenization (FBT) for data pre-processing on 1-D Convolutional Neural Network (CNN) model. We found that FBT gives a shorter input dimension reducing the training time without affecting the sensitivity and specificity of classification. We employed the deep learning techniques, mainly CNN and recurrent neural network with Long Short Term Memory (LSTM) and random forest (RF) classifier for promoter classification at k-mer sizes of 2, 4 and 8. We found CNN to be superior in classification of promoters from non-promoter sequences (binary classification) as well as species-specific classification of promoter sequences (multiclass classification). In summary, the contribution of this work lies in the use of synthetic shuffled negative dataset and frequency-based tokenization for pre-processing. This study provides a comprehensive and generic framework for classification tasks in genomic applications and can be extended to various classification problems.
Spear Phishing is a harmful cyber-attack facing business and individuals worldwide. Considerable research has been conducted recently into the use of Machine Learning (ML) techniques to detect spear-phishing emails. ML-based solutions may suffer from zero-day attacks; unseen attacks unaccounted for in the training data. As new attacks emerge, classifiers trained on older data are unable to detect these new varieties of attacks resulting in increasingly inaccurate predictions. Spear Phishing detection also faces scalability challenges due to the growth of the required features which is proportional to the number of the senders within a receiver mailbox. This differs from traditional phishing attacks which typically perform only a binary classification between phishing and benign emails. Therefore, we devise a possible solution to these problems, named RAIDER: Reinforcement AIded Spear Phishing DEtectoR. A reinforcement-learning based feature evaluation system that can automatically find the optimum features for detecting different types of attacks. By leveraging a reward and penalty system, RAIDER allows for autonomous features selection. RAIDER also keeps the number of features to a minimum by selecting only the significant features to represent phishing emails and detect spear-phishing attacks. After extensive evaluation of RAIDER over 11,000 emails and across 3 attack scenarios, our results suggest that using reinforcement learning to automatically identify the significant features could reduce the dimensions of the required features by 55% in comparison to existing ML-based systems. It also improves the accuracy of detecting spoofing attacks by 4% from 90% to 94%. In addition, RAIDER demonstrates reasonable detection accuracy even against a sophisticated attack named Known Sender in which spear-phishing emails greatly resemble those of the impersonated sender.
The World Wide Web is not only one of the most important platforms of communication and information at present, but also an area of growing interest for scientific research. This motivates a lot of work and projects that require large amounts of data. However, there is no dataset that integrates the parameters and visual appearance of Web pages, because its collection is a costly task in terms of time and effort. With the support of various computer tools and programming scripts, we have created a large dataset of 49,438 Web pages. It consists of visual, textual and numerical data types, includes all countries worldwide, and considers a broad range of topics such as art, entertainment, economy, business, education, government, news, media, science, and environment, covering different cultural characteristics and varied design preferences. In this paper, we describe the process of collecting, debugging and publishing the final product, which is freely available. To demonstrate the usefulness of our dataset, we expose a binary classification model for detecting error Web pages, and a multi-class Web subject-based categorization, both problems using convolutional neural networks.
We present DaLAJ 1.0, a Dataset for Linguistic Acceptability Judgments for Swedish, comprising 9 596 sentences in its first version; and the initial experiment using it for the binary classification task. DaLAJ is based on the SweLL second language learner data, consisting of essays at different levels of proficiency. To make sure the dataset can be freely available despite the GDPR regulations, we have sentence-scrambled learner essays and removed part of the metadata about learners, keeping for each sentence only information about the mother tongue and the level of the course where the essay has been written. We use the normalized version of learner language as the basis for the DaLAJ sentences, and keep only one error per sentence. We repeat the same sentence for each individual correction tag used in the sentence. For DaLAJ 1.0 we have used four error categories (out of 35 available in SweLL), all connected to lexical or word-building choices. Our baseline results for the binary classification show an accuracy of 58% for DaLAJ 1.0 using BERT embeddings. The dataset is included in the SwedishGlue (Swe. SuperLim) benchmark. Below, we describe the format of the dataset, first experiments, our insights and the motivation for the chosen approach to data sharing.
We consider active learning for binary classification in the agnostic pool-based setting. The vast majority of works in active learning in the agnostic setting are inspired by the CAL algorithm where each query is uniformly sampled from the disagreement region of the current version space. The sample complexity of such algorithms is described by a quantity known as the disagreement coefficient which captures both the geometry of the hypothesis space as well as the underlying probability space. To date, the disagreement coefficient has been justified by minimax lower bounds only, leaving the door open for superior instance dependent sample complexities. In this work we propose an algorithm that, in contrast to uniform sampling over the disagreement region, solves an experimental design problem to determine a distribution over examples from which to request labels. We show that the new approach achieves sample complexity bounds that are never worse than the best disagreement coefficient-based bounds, but in specific cases can be dramatically smaller. From a practical perspective, the proposed algorithm requires no hyperparameters to tune (e.g., to control the aggressiveness of sampling), and is computationally efficient by means of assuming access to an empirical risk minimization oracle (without any constraints). Empirically, we demonstrate that our algorithm is superior to state of the art agnostic active learning algorithms on image classification datasets.
Patient monitoring is vital in all stages of care. We here report the development and validation of ICU length of stay and mortality prediction models. The models will be used in an intelligent ICU patient monitoring module of an Intelligent Remote Patient Monitoring (IRPM) framework that monitors the health status of patients, and generates timely alerts, maneuver guidance, or reports when adverse medical conditions are predicted. We utilized the publicly available Medical Information Mart for Intensive Care (MIMIC) database to extract ICU stay data for adult patients to build two prediction models: one for mortality prediction and another for ICU length of stay. For the mortality model, we applied six commonly used machine learning (ML) binary classification algorithms for predicting the discharge status (survived or not). For the length of stay model, we applied the same six ML algorithms for binary classification using the median patient population ICU stay of 2.64 days. For the regression-based classification, we used two ML algorithms for predicting the number of days. We built two variations of each prediction model: one using 12 baseline demographic and vital sign features, and the other based on our proposed quantiles approach, in which we use 21 extra features engineered from the baseline vital sign features, including their modified means, standard deviations, and quantile percentages. We could perform predictive modeling with minimal features while maintaining reasonable performance using the quantiles approach. The best accuracy achieved in the mortality model was approximately 89% using the random forest algorithm. The highest accuracy achieved in the length of stay model, based on the population median ICU stay (2.64 days), was approximately 65% using the random forest algorithm.
Recommendation systems have been extensively studied by many literature in the past and are ubiquitous in online advertisement, shopping industry/e-commerce, query suggestions in search engines, and friend recommendation in social networks. Moreover, restaurant/music/product/movie/news/app recommendations are only a few of the applications of a recommender system. A small percent improvement on the CTR prediction accuracy has been mentioned to add millions of dollars of revenue to the advertisement industry. Click-Through-Rate (CTR) prediction is a special version of recommender system in which the goal is predicting whether or not a user is going to click on a recommended item. A content-based recommendation approach takes into account the past history of the user's behavior, i.e. the recommended products and the users reaction to them. So, a personalized model that recommends the right item to the right user at the right time is the key to building such a model. On the other hand, the so-called collaborative filtering approach incorporates the click history of the users who are very similar to a particular user, thereby helping the recommender to come up with a more confident prediction for that particular user by leveraging the wider knowledge of users who share their taste in a connected network of users. In this project, we are interested in building a CTR predictor using Graph Neural Networks complemented by an online learning algorithm that models such dynamic interactions. By framing the problem as a binary classification task, we have evaluated this system both on the offline models (GNN, Deep Factorization Machines) with test-AUC of 0.7417 and on the online learning model with test-AUC of 0.7585 using a sub-sampled version of Criteo public dataset consisting of 10,000 data points.
The task of morphological classification is complex for simple parameterization, but important for research in the galaxy evolution field. Future galaxy surveys (e.g. EUCLID) will collect data about more than a $10^9$ galaxies. To obtain morphological information one needs to involve people to mark up galaxy images, which requires either a considerable amount of money or a huge number of volunteers. We propose an effective semi-supervised approach for galaxy morphology classification task, based on active learning of adversarial autoencoder (AAE) model. For a binary classification problem (top level question of Galaxy Zoo 2 decision tree) we achieved accuracy 93.1% on the test part with only 0.86 millions markup actions, this model can easily scale up on any number of images. Our best model with additional markup achieves accuracy of 95.5%. To the best of our knowledge it is a first time AAE semi-supervised learning model used in astronomy.