Covariance matrices of noisy multichannel electroencephalogram time series data are hard to estimate due to high dimensionality. In brain-computer interfaces (BCI) based on event-related potentials and a linear discriminant analysis (LDA) for classification, the state of the art to address this problem is by shrinkage regularization. We propose a novel idea to tackle this problem by enforcing a block-Toeplitz structure for the covariance matrix of the LDA, which implements an assumption of signal stationarity in short time windows for each channel. On data of 213 subjects collected under 13 event-related potential BCI protocols, the resulting 'ToeplitzLDA' significantly increases the binary classification performance compared to shrinkage regularized LDA (up to 6 AUC points) and Riemannian classification approaches (up to 2 AUC points). This translates to greatly improved application level performances, as exemplified on data recorded during an unsupervised visual speller application, where spelling errors could be reduced by 81% on average for 25 subjects. Aside from lower memory and time complexity for LDA training, ToeplitzLDA proved to be almost invariant even to a twenty-fold time dimensionality enlargement, which reduces the need of expert knowledge regarding feature extraction.
We propose Distribution Embedding Networks (DEN) for classification with small data. In the same spirit of meta-learning, DEN learns from a diverse set of training tasks with the goal to generalize to unseen target tasks. Unlike existing approaches which require the inputs of training and target tasks to have the same dimension with possibly similar distributions, DEN allows training and target tasks to live in heterogeneous input spaces. This is especially useful for tabular-data tasks where labeled data from related tasks are scarce. DEN uses a three-block architecture: a covariate transformation block followed by a distribution embedding block and then a classification block. We provide theoretical insights to show that this architecture allows the embedding and classification blocks to be fixed after pre-training on a diverse set of tasks; only the covariate transformation block with relatively few parameters needs to be fine-tuned for each new task. To facilitate training, we also propose an approach to synthesize binary classification tasks, and demonstrate that DEN outperforms existing methods in a number of synthetic and real tasks in numerical studies.
Algorithms and Machine Learning (ML) are increasingly affecting everyday life and several decision-making processes, where ML has an advantage due to scalability or superior performance. Fairness in such applications is crucial, where models should not discriminate their results based on race, gender, or other protected groups. This is especially crucial for models affecting very sensitive topics, like interview invitation or recidivism prediction. Fairness is not commonly studied for regression problems compared to binary classification problems; hence, we present a simple, yet effective method based on normalisation (FaiReg), which minimises the impact of unfairness in regression problems, especially due to labelling bias. We present a theoretical analysis of the method, in addition to an empirical comparison against two standard methods for fairness, namely data balancing and adversarial training. We also include a hybrid formulation (FaiRegH), merging the presented method with data balancing, in an attempt to face labelling and sampling biases simultaneously. The experiments are conducted on the multimodal dataset First Impressions (FI) with various labels, namely Big-Five personality prediction and interview screening score. The results show the superior performance of diminishing the effects of unfairness better than data balancing, also without deteriorating the performance of the original problem as much as adversarial training. Fairness is evaluated based on the Equal Accuracy (EA) and Statistical Parity (SP) constraints. The experiments present a setup that enhances the fairness for several protected variables simultaneously.
There is a fundamental limitation in the prediction performance that a machine learning model can achieve due to the inevitable uncertainty of the prediction target. In classification problems, this can be characterized by the Bayes error, which is the best achievable error with any classifier. The Bayes error can be used as a criterion to evaluate classifiers with state-of-the-art performance and can be used to detect test set overfitting. We propose a simple and direct Bayes error estimator, where we just take the mean of the labels that show \emph{uncertainty} of the class assignments. Our flexible approach enables us to perform Bayes error estimation even for weakly supervised data. In contrast to others, our method is model-free and even instance-free. Moreover, it has no hyperparameters and gives a more accurate estimate of the Bayes error than several baselines empirically. Experiments using our method suggest that recently proposed deep networks such as the Vision Transformer may have reached, or is about to reach, the Bayes error for benchmark datasets. Finally, we discuss how we can study the inherent difficulty of the acceptance/rejection decision for scientific articles, by estimating the Bayes error of the ICLR papers from 2017 to 2023.
Device fingerprinting combined with Machine and Deep Learning (ML/DL) report promising performance when detecting cyberattacks targeting data managed by resource-constrained spectrum sensors. However, the amount of data needed to train models and the privacy concerns of such scenarios limit the applicability of centralized ML/DL-based approaches. Federated learning (FL) addresses these limitations by creating federated and privacy-preserving models. However, FL is vulnerable to malicious participants, and the impact of adversarial attacks on federated models detecting spectrum sensing data falsification (SSDF) attacks on spectrum sensors has not been studied. To address this challenge, the first contribution of this work is the creation of a novel dataset suitable for FL and modeling the behavior (usage of CPU, memory, or file system, among others) of resource-constrained spectrum sensors affected by different SSDF attacks. The second contribution is a pool of experiments analyzing and comparing the robustness of federated models according to i) three families of spectrum sensors, ii) eight SSDF attacks, iii) four scenarios dealing with unsupervised (anomaly detection) and supervised (binary classification) federated models, iv) up to 33% of malicious participants implementing data and model poisoning attacks, and v) four aggregation functions acting as anti-adversarial mechanisms to increase the models robustness.
How do two distributions of texts differ? Humans are slow at answering this, since discovering patterns might require tediously reading through hundreds of samples. We propose to automatically summarize the differences by "learning a natural language hypothesis": given two distributions $D_{0}$ and $D_{1}$, we search for a description that is more often true for $D_{1}$, e.g., "is military-related." To tackle this problem, we fine-tune GPT-3 to propose descriptions with the prompt: "[samples of $D_{0}$] + [samples of $D_{1}$] + the difference between them is_____." We then re-rank the descriptions by checking how often they hold on a larger set of samples with a learned verifier. On a benchmark of 54 real-world binary classification tasks, while GPT-3 Curie (13B) only generates a description similar to human annotation 7% of the time, the performance reaches 61% with fine-tuning and re-ranking, and our best system using GPT-3 Davinci (175B) reaches 76%. We apply our system to describe distribution shifts, debug dataset shortcuts, summarize unknown tasks, and label text clusters, and present analyses based on automatically generated descriptions.
Despite the introduction of vaccines, Coronavirus disease (COVID-19) remains a worldwide dilemma, continuously developing new variants such as Delta and the recent Omicron. The current standard for testing is through polymerase chain reaction (PCR). However, PCRs can be expensive, slow, and/or inaccessible to many people. X-rays on the other hand have been readily used since the early 20th century and are relatively cheaper, quicker to obtain, and typically covered by health insurance. With a careful selection of model, hyperparameters, and augmentations, we show that it is possible to develop models with 83% accuracy in binary classification and 64% in multi-class for detecting COVID-19 infections from chest x-rays.
Facial images disclose many hidden personal traits such as age, gender, race, health, emotion, and psychology. Understanding these traits will help to classify the people in different attributes. In this paper, we have presented a novel method for classifying images using a pretrained transformer model. We apply the pretrained transformer for the binary classification of facial images in criminal and non-criminal classes. The pretrained transformer of GPT-2 is trained to generate text and then fine-tuned to classify facial images. During the finetuning process with images, most of the layers of GT-2 are frozen during backpropagation and the model is frozen pretrained transformer (FPT). The FPT acts as a universal image classifier, and this paper shows the application of FPT on facial images. We also use our FPT on encrypted images for classification. Our FPT shows high accuracy on both raw facial images and encrypted images. We hypothesize the meta-learning capacity FPT gained because of its large size and trained on a large size with theory and experiments. The GPT-2 trained to generate a single word token at a time, through the autoregressive process, forced to heavy-tail distribution. Then the FPT uses the heavy-tail property as its meta-learning capacity for classifying images. Our work shows one way to avoid bias during the machine classification of images.The FPT encodes worldly knowledge because of the pretraining of one text, which it uses during the classification. The statistical error of classification is reduced because of the added context gained from the text.Our paper shows the ethical dimension of using encrypted data for classification.Criminal images are sensitive to share across the boundary but encrypted largely evades ethical concern.FPT showing good classification accuracy on encrypted images shows promise for further research on privacy-preserving machine learning.
We introduce Explanatory Learning (EL), a framework to let machines use existing knowledge buried in symbolic sequences -- e.g. explanations written in hieroglyphic -- by autonomously learning to interpret them. In EL, the burden of interpreting symbols is not left to humans or rigid human-coded compilers, as done in Program Synthesis. Rather, EL calls for a learned interpreter, built upon a limited collection of symbolic sequences paired with observations of several phenomena. This interpreter can be used to make predictions on a novel phenomenon given its explanation, and even to find that explanation using only a handful of observations, like human scientists do. We formulate the EL problem as a simple binary classification task, so that common end-to-end approaches aligned with the dominant empiricist view of machine learning could, in principle, solve it. To these models, we oppose Critical Rationalist Networks (CRNs), which instead embrace a rationalist view on the acquisition of knowledge. CRNs express several desired properties by construction, they are truly explainable, can adjust their processing at test-time for harder inferences, and can offer strong confidence guarantees on their predictions. As a final contribution, we introduce Odeen, a basic EL environment that simulates a small flatland-style universe full of phenomena to explain. Using Odeen as a testbed, we show how CRNs outperform empiricist end-to-end approaches of similar size and architecture (Transformers) in discovering explanations for novel phenomena.
Automatic classification of diabetic retinopathy from retinal images has been widely studied using deep neural networks with impressive results. However, there is a clinical need for estimation of the uncertainty in the classifications, a shortcoming of modern neural networks. Recently, approximate Bayesian deep learning methods have been proposed for the task but the studies have only considered the binary referable/non-referable diabetic retinopathy classification applied to benchmark datasets. We present novel results by systematically investigating a clinical dataset and a clinically relevant 5-class classification scheme, in addition to benchmark datasets and the binary classification scheme. Moreover, we derive a connection between uncertainty measures and classifier risk, from which we develop a new uncertainty measure. We observe that the previously proposed entropy-based uncertainty measure generalizes to the clinical dataset on the binary classification scheme but not on the 5-class scheme, whereas our new uncertainty measure generalizes to the latter case.
Balancing the data before training a classifier is a popular technique to address the challenges of imbalanced binary classification in tabular data. Balancing is commonly achieved by duplication of minority samples or by generation of synthetic minority samples. While it is well known that balancing affects each classifier differently, most prior empirical studies did not include strong state-of-the-art (SOTA) classifiers as baselines. In this work, we are interested in understanding whether balancing is beneficial, particularly in the context of SOTA classifiers. Thus, we conduct extensive experiments considering three SOTA classifiers along the weaker learners used in previous investigations. Additionally, we carefully discern proper metrics, consistent and non-consistent algorithms and hyper-parameter selection methods and show that these have a significant impact on prediction quality and on the effectiveness of balancing. Our results support the known utility of balancing for weak classifiers. However, we find that balancing does not improve prediction performance for the strong ones. We further identify several other scenarios for which balancing is effective and observe that prior studies demonstrated the utility of balancing by focusing on these settings.
Performance forecasting is an age-old problem in economics and finance. Recently, developments in machine learning and neural networks have given rise to non-linear time series models that provide modern and promising alternatives to traditional methods of analysis. In this paper, we present an ensemble of independent and parallel long short-term memory (LSTM) neural networks for the prediction of stock price movement. LSTMs have been shown to be especially suited for time series data due to their ability to incorporate past information, while neural network ensembles have been found to reduce variability in results and improve generalization. A binary classification problem based on the median of returns is used, and the ensemble's forecast depends on a threshold value, which is the minimum number of LSTMs required to agree upon the result. The model is applied to the constituents of the smaller, less efficient Stockholm OMX30 instead of other major market indices such as the DJIA and S&P500 commonly found in literature. With a straightforward trading strategy, comparisons with a randomly chosen portfolio and a portfolio containing all the stocks in the index show that the portfolio resulting from the LSTM ensemble provides better average daily returns and higher cumulative returns over time. Moreover, the LSTM portfolio also exhibits less volatility, leading to higher risk-return ratios.
With the fast-growing number of classification models being produced every day, numerous model interpretation and comparison solutions have also been introduced. For example, LIME and SHAP can interpret what input features contribute more to a classifier's output predictions. Different numerical metrics (e.g., accuracy) can be used to easily compare two classifiers. However, few works can interpret the contribution of a data feature to a classifier in comparison with its contribution to another classifier. This comparative interpretation can help to disclose the fundamental difference between two classifiers, select classifiers in different feature conditions, and better ensemble two classifiers. To accomplish it, we propose a learning-from-disagreement (LFD) framework to visually compare two classification models. Specifically, LFD identifies data instances with disagreed predictions from two compared classifiers and trains a discriminator to learn from the disagreed instances. As the two classifiers' training features may not be available, we train the discriminator through a set of meta-features proposed based on certain hypotheses of the classifiers to probe their behaviors. Interpreting the trained discriminator with the SHAP values of different meta-features, we provide actionable insights into the compared classifiers. Also, we introduce multiple metrics to profile the importance of meta-features from different perspectives. With these metrics, one can easily identify meta-features with the most complementary behaviors in two classifiers, and use them to better ensemble the classifiers. We focus on binary classification models in the financial services and advertising industry to demonstrate the efficacy of our proposed framework and visualizations.
We propose the task of automatically identifying papers used as baselines in a scientific article. We frame the problem as a binary classification task where all the references in a paper are to be classified as either baselines or non-baselines. This is a challenging problem due to the numerous ways in which a baseline reference can appear in a paper. We develop a dataset of $2,075$ papers from ACL anthology corpus with all their references manually annotated as one of the two classes. We develop a multi-module attention-based neural classifier for the baseline classification task that outperforms four state-of-the-art citation role classification methods when applied to the baseline classification task. We also present an analysis of the errors made by the proposed classifier, eliciting the challenges that make baseline identification a challenging problem.
Face Anti Spoofing (FAS) systems are used to identify malicious spoofing attempts targeting face recognition systems using mediums such as video replay or printed papers. With increasing adoption of face recognition technology as a biometric authentication method, FAS techniques are gaining in importance. From a learning perspective, such systems pose a binary classification task. When implemented with Neural Network based solutions, it is common to use the binary cross entropy (BCE) function as the loss to optimize. In this study, we propose a variant of BCE that enforces a margin in angular space and incorporate it in training the DeepPixBis model [1]. In addition, we also present a method to incorporate such a loss for attentive pixel wise supervision applicable in a fully convolutional setting. Our proposed approach achieves competitive scores in both intra and inter-dataset testing on multiple benchmark datasets, consistently outperforming vanilla DeepPixBis. Interestingly, in the case of Protocol 4 of OULU-NPU, considered to be the hardest protocol, our proposed method achieves 5.22% ACER, which is only 0.22% higher than the current State of the Art without requiring any expensive Neural Architecture Search.
Positive-unlabeled learning (PU learning) is known as a special case of semi-supervised binary classification where only a fraction of positive examples are labeled. The challenge is then to find the correct classifier despite this lack of information. Recently, new methodologies have been introduced to address the case where the probability of being labeled may depend on the covariates. In this paper, we are interested in establishing risk bounds for PU learning under this general assumption. In addition, we quantify the impact of label noise on PU learning compared to standard classification setting. Finally, we provide a lower bound on minimax risk proving that the upper bound is almost optimal.
Signed link prediction in graphs is an important problem that has applications in diverse domains. It is a binary classification problem that predicts whether an edge between a pair of nodes is positive or negative. Existing approaches for link prediction in unsigned networks cannot be directly applied for signed link prediction due to their inherent differences. Further, additional structural constraints, like, the structural balance property of the signed networks must be considered for signed link prediction. Recent signed link prediction approaches generate node representations using either generative models or discriminative models. Inspired by the recent success of Generative Adversarial Network (GAN) based models which comprises of a discriminator and generator in several applications, we propose a Generative Adversarial Network (GAN) based model for signed networks, SigGAN. It considers the requirements of signed networks, such as, integration of information from negative edges, high imbalance in number of positive and negative edges and structural balance theory. Comparing the performance with state of the art techniques on several real-world datasets validates the effectiveness of SigGAN.
A key component of a quantum machine learning model operating on classical inputs is the design of an embedding circuit mapping inputs to a quantum state. This paper studies a transfer learning setting in which classical-to-quantum embedding is carried out by an arbitrary parametric quantum circuit that is pre-trained based on data from a source task. At run time, a binary quantum classifier of the embedding is optimized based on data from the target task of interest. The average excess risk, i.e., the optimality gap, of the resulting classifier depends on how (dis)similar the source and target tasks are. We introduce a new measure of (dis)similarity between the binary quantum classification tasks via the trace distances. An upper bound on the optimality gap is derived in terms of the proposed task (dis)similarity measure, two R$\'e$nyi mutual information terms between classical input and quantum embedding under source and target tasks, as well as a measure of complexity of the combined space of quantum embeddings and classifiers under the source task. The theoretical results are validated on a simple binary classification example.
Parkinsons disease, PD, is a chronic condition that affects motor skills and includes symptoms like tremors and rigidity. The current diagnostic procedure uses patient assessments to evaluate symptoms and sometimes a magnetic resonance imaging or MRI scan. However, symptom variations cause inaccurate assessments, and the analysis of MRI scans requires experienced specialists. This research proposes to accurately diagnose PD severity with deep learning by combining symptoms data and MRI data from the Parkinsons Progression Markers Initiative database. A new hybrid model architecture was implemented to fully utilize both forms of clinical data, and models based on only symptoms and only MRI scans were also developed. The symptoms based model integrates a fully connected deep learning neural network, and the MRI scans and hybrid models integrate transfer learning based convolutional neural networks. Instead of performing only binary classification, all models diagnose patients into five severity categories, with stage zero representing healthy patients and stages four and five representing patients with PD. The symptoms only, MRI scans only, and hybrid models achieved accuracies of 0.77, 0.68, and 0.94, respectively. The hybrid model also had high precision and recall scores of 0.94 and 0.95. Real clinical cases confirm the strong performance of the hybrid, where patients were classified incorrectly with both other models but correctly by the hybrid. It is also consistent across the five severity stages, indicating accurate early detection. This is the first report to combine symptoms data and MRI scans with a machine learning approach on such a large scale.
Applying standard machine learning approaches for classification can produce unequal results across different demographic groups. When then used in real-world settings, these inequities can have negative societal impacts. This has motivated the development of various approaches to fair classification with machine learning models in recent years. In this paper, we consider the problem of modifying the predictions of a blackbox machine learning classifier in order to achieve fairness in a multiclass setting. To accomplish this, we extend the 'post-processing' approach in Hardt et al. 2016, which focuses on fairness for binary classification, to the setting of fair multiclass classification. We explore when our approach produces both fair and accurate predictions through systematic synthetic experiments and also evaluate discrimination-fairness tradeoffs on several publicly available real-world application datasets. We find that overall, our approach produces minor drops in accuracy and enforces fairness when the number of individuals in the dataset is high relative to the number of classes and protected groups.