Alzheimer's Dementia (AD) is an incurable, debilitating, and progressive neurodegenerative condition that affects cognitive function. Early diagnosis is important as therapeutics can delay progression and give those diagnosed vital time. Developing models that analyse spontaneous speech could eventually provide an efficient diagnostic modality for earlier diagnosis of AD. The Alzheimer's Dementia Recognition through Spontaneous Speech task offers acoustically pre-processed and balanced datasets for the classification and prediction of AD and associated phenotypes through the modelling of spontaneous speech. We exclusively analyse the supplied textual transcripts of the spontaneous speech dataset, building and comparing performance across numerous models for the classification of AD vs controls and the prediction of Mental Mini State Exam scores. We rigorously train and evaluate Support Vector Machines (SVMs), Gradient Boosting Decision Trees (GBDT), and Conditional Random Fields (CRFs) alongside deep learning Transformer based models. We find our top performing models to be a simple Term Frequency-Inverse Document Frequency (TF-IDF) vectoriser as input into a SVM model and a pre-trained Transformer based model `DistilBERT' when used as an embedding layer into simple linear models. We demonstrate test set scores of 0.81-0.82 across classification metrics and a RMSE of 4.58.
In recent decades, the automatic video surveillance system has gained significant importance in computer vision community. The crucial objective of surveillance is monitoring and security in public places. In the traditional Local Binary Pattern, the feature description is somehow inaccurate, and the feature size is large enough. Therefore, to overcome these shortcomings, our research proposed a detection algorithm for a human with or without carrying baggage. The Local tri-directional pattern descriptor is exhibited to extract features of different human body parts including head, trunk, and limbs. Then with the help of support vector machine, extracted features are trained and evaluated. Experimental results on INRIA and MSMT17 V1 datasets show that LtriDP outperforms several state-of-the-art feature descriptors and validate its effectiveness.
Modern neural architectures for classification tasks are trained using the cross-entropy loss, which is widely believed to be empirically superior to the square loss. In this work we provide evidence indicating that this belief may not be well-founded. We explore several major neural architectures and a range of standard benchmark datasets for NLP, automatic speech recognition (ASR) and computer vision tasks to show that these architectures, with the same hyper-parameter settings as reported in the literature, perform comparably or better when trained with the square loss, even after equalizing computational resources. Indeed, we observe that the square loss produces better results in the dominant majority of NLP and ASR experiments. Cross-entropy appears to have a slight edge on computer vision tasks. We argue that there is little compelling empirical or theoretical evidence indicating a clear-cut advantage to the cross-entropy loss. Indeed, in our experiments, performance on nearly all non-vision tasks can be improved, sometimes significantly, by switching to the square loss. Furthermore, training with square loss appears to be less sensitive to the randomness in initialization. We posit that training using the square loss for classification needs to be a part of best practices of modern deep learning on equal footing with cross-entropy.
Image classification is central to the big data revolution in medicine. Improved information processing methods for diagnosis and classification of digital medical images have shown to be successful via deep learning approaches. As this field is explored, there are limitations to the performance of traditional supervised classifiers. This paper outlines an approach that is different from the current medical image classification tasks that view the issue as multi-class classification. We performed a hierarchical classification using our Hierarchical Medical Image classification (HMIC) approach. HMIC uses stacks of deep learning models to give particular comprehension at each level of the clinical picture hierarchy. For testing our performance, we use biopsy of the small bowel images that contain three categories in the parent level (Celiac Disease, Environmental Enteropathy, and histologically normal controls). For the child level, Celiac Disease Severity is classified into 4 classes (I, IIIa, IIIb, and IIIC).
Many important questions about a model cannot be answered just by explaining how much each feature contributes to its output. To answer a broader set of questions, we generalize a popular, mathematically well-grounded explanation technique, Shapley Additive Explanations (SHAP). Our new method - Generalized Shapley Additive Explanations (G-SHAP) - produces many additional types of explanations, including: 1) General classification explanations; Why is this sample more likely to belong to one class rather than another? 2) Intergroup differences; Why do our model's predictions differ between groups of observations? 3) Model failure; Why does our model perform poorly on a given sample? We formally define these types of explanations and illustrate their practical use on real data.
Adversarial attacks have been widely studied for general classification tasks, but remain unexplored in the context of fine-grained recognition, where the inter-class similarities facilitate the attacker's task. In this paper, we identify the proximity of the latent representations of different classes in fine-grained recognition networks as a key factor to the success of adversarial attacks. We therefore introduce an attention-based regularization mechanism that maximally separates the discriminative latent features of different classes while minimizing the contribution of the non-discriminative regions to the final class prediction. As evidenced by our experiments, this allows us to significantly improve robustness to adversarial attacks, to the point of matching or even surpassing that of adversarial training, but without requiring access to adversarial samples.
In classical statistical learning theory, one of the most well studied problems is that of binary classification. The information-theoretic sample complexity of this task is tightly characterized by the Vapnik-Chervonenkis (VC) dimension. A quantum analog of this task, with training data given as a quantum state has also been intensely studied and is now known to have the same sample complexity as its classical counterpart. We propose a novel quantum version of the classical binary classification task by considering maps with classical input and quantum output and corresponding classical-quantum training data. We discuss learning strategies for the agnostic and for the realizable case and study their performance to obtain sample complexity upper bounds. Moreover, we provide sample complexity lower bounds which show that our upper bounds are essentially tight for pure output states. In particular, we see that the sample complexity is the same as in the classical binary classification task w.r.t. its dependence on accuracy, confidence and the VC-dimension.
The use of quadratic discriminant analysis (QDA) or its regularized version (R-QDA) for classification is often not recommended, due to its well-acknowledged high sensitivity to the estimation noise of the covariance matrix. This becomes all the more the case in unbalanced data settings for which it has been found that R-QDA becomes equivalent to the classifier that assigns all observations to the same class. In this paper, we propose an improved R-QDA that is based on the use of two regularization parameters and a modified bias, properly chosen to avoid inappropriate behaviors of R-QDA in unbalanced settings and to ensure the best possible classification performance. The design of the proposed classifier builds on a refined asymptotic analysis of its performance when the number of samples and that of features grow large simultaneously, which allows to cope efficiently with the high-dimensionality frequently met within the big data paradigm. The performance of the proposed classifier is assessed on both real and synthetic data sets and was shown to be much better than what one would expect from a traditional R-QDA.
In the advent of wearable body-cameras, human activity classification from First-Person Videos (FPV) has become a topic of increasing importance for various applications, including in life-logging, law-enforcement, sports, workplace, and healthcare. One of the challenging aspects of FPV is its exposure to potentially sensitive objects within the user's field of view. In this work, we developed a privacy-aware activity classification system focusing on office videos. We utilized a Mask-RCNN with an Inception-ResNet hybrid as a feature extractor for detecting, and then blurring out sensitive objects (e.g., digital screens, human face, paper) from the videos. For activity classification, we incorporate an ensemble of Recurrent Neural Networks (RNNs) with ResNet, ResNext, and DenseNet based feature extractors. The proposed system was trained and evaluated on the FPV office video dataset that includes 18-classes made available through the IEEE Video and Image Processing (VIP) Cup 2019 competition. On the original unprotected FPVs, the proposed activity classifier ensemble reached an accuracy of 85.078% with precision, recall, and F1 scores of 0.88, 0.85 & 0.86, respectively. On privacy protected videos, the performances were slightly degraded, with accuracy, precision, recall, and F1 scores at 73.68%, 0.79, 0.75, and 0.74, respectively. The presented system won the 3rd prize in the IEEE VIP Cup 2019 competition.
While much work on deep latent variable models of text uses continuous latent variables, discrete latent variables are interesting because they are more interpretable and typically more space efficient. We consider several approaches to learning discrete latent variable models for text in the case where exact marginalization over these variables is intractable. We compare the performance of the learned representations as features for low-resource document and sentence classification. Our best models outperform the previous best reported results with continuous representations in these low-resource settings, while learning significantly more compressed representations. Interestingly, we find that an amortized variant of Hard EM performs particularly well in the lowest-resource regimes.
Similarity learning is a general problem to elicit useful representations by predicting the relationship between a pair of patterns. This problem is related to various important preprocessing tasks such as metric learning, kernel learning, and contrastive learning. A classifier built upon the representations is expected to perform well in downstream classification; however, little theory has been given in literature so far and thereby the relationship between similarity and classification has remained elusive. Therefore, we tackle a fundamental question: can similarity information provably leads a model to perform well in downstream classification? In this paper, we reveal that a product-type formulation of similarity learning is strongly related to an objective of binary classification. We further show that these two different problems are explicitly connected by an excess risk bound. Consequently, our results elucidate that similarity learning is capable of solving binary classification by directly eliciting a decision boundary.
We analyze in a closed form the learning dynamics of stochastic gradient descent (SGD) for a single-layer neural network classifying a high-dimensional Gaussian mixture where each cluster is assigned one of two labels. This problem provides a prototype of a non-convex loss landscape with interpolating regimes and a large generalization gap. We define a particular stochastic process for which SGD can be extended to a continuous-time limit that we call stochastic gradient flow. In the full-batch limit, we recover the standard gradient flow. We apply dynamical mean-field theory from statistical physics to track the dynamics of the algorithm in the high-dimensional limit via a self-consistent stochastic process. We explore the performance of the algorithm as a function of the control parameters shedding light on how it navigates the loss landscape.
We develop Distributionally Robust Optimization (DRO) formulations for Multivariate Linear Regression (MLR) and Multiclass Logistic Regression (MLG) when both the covariates and responses/labels may be contaminated by outliers. The DRO framework uses a probabilistic ambiguity set defined as a ball of distributions that are close to the empirical distribution of the training set in the sense of the Wasserstein metric. We relax the DRO formulation into a regularized learning problem whose regularizer is a norm of the coefficient matrix. We establish out-of-sample performance guarantees for the solutions to our model, offering insights on the role of the regularizer in controlling the prediction error. Experimental results show that our approach improves the predictive error by 7% -- 37% for MLR, and a metric of robustness by 100% for MLG.
Detecting malicious intent behavior such as sharing hate speech has become an important challenge for social networking platforms. The method of automated hate speech detection for social media posts is often challenged by the complexity of capturing the context of the user expression with potential hate intent. We hypothesize that semantic features can help enrich the context representation of word senses in a social media post for machine learning algorithms. This paper presents a novel empirical study of diverse semantic features for hate speech classification task on social media posts. Specifically, we present an extensive empirical analysis, where we test the features of vector space model representation for corpus-based semantics, neural word embedding representation for distributional semantics, and declarative knowledge patterns from external knowledge base for domain semantics. Our experimental results show that ensembling the diverse feature representations improves the efficiency of hateful behavior classification in contrast to the case of a single type of feature representation. Results on two popular Twitter datasets for the hate speech detection task showed a consistent performance gain for the classification models that were based on the hybrid feature representation (absolute gain in F1 score up to 3.0%). The application of the proposed method of combining diverse feature representations can help in improving social media analytics systems for monitoring human behavior.
The de-facto approach to many vision tasks is to start from pretrained visual representations, typically learned via supervised training on ImageNet. Recent methods have explored unsupervised pretraining to scale to vast quantities of unlabeled images. In contrast, we aim to learn high-quality visual representations from fewer images. To this end, we revisit supervised pretraining, and seek data-efficient alternatives to classification-based pretraining. We propose VirTex -- a pretraining approach using semantically dense captions to learn visual representations. We train convolutional networks from scratch on COCO Captions, and transfer them to downstream recognition tasks including image classification, object detection, and instance segmentation. On all tasks, VirTex yields features that match or exceed those learned on ImageNet -- supervised or unsupervised -- despite using up to ten times fewer images.
In this paper, we show that although the minimizers of cross-entropy and related classification losses are off at infinity, network weights learned by gradient flow converge in direction, with an immediate corollary that network predictions, training errors, and the margin distribution also converge. This proof holds for deep homogeneous networks -- a broad class of networks allowing for ReLU, max-pooling, linear, and convolutional layers -- and we additionally provide empirical support not just close to the theory (e.g., the AlexNet), but also on non-homogeneous networks (e.g., the DenseNet). If the network further has locally Lipschitz gradients, we show that these gradients also converge in direction, and asymptotically align with the gradient flow path, with consequences on margin maximization, convergence of saliency maps, and a few other settings. Our analysis complements and is distinct from the well-known neural tangent and mean-field theories, and in particular makes no requirements on network width and initialization, instead merely requiring perfect classification accuracy. The proof proceeds by developing a theory of unbounded nonsmooth Kurdyka-{\L}ojasiewicz inequalities for functions definable in an o-minimal structure, and is also applicable outside deep learning.
Region sampling or weighting is significantly important to the success of modern region-based object detectors. Unlike some previous works, which only focus on "hard" samples when optimizing the objective function, we argue that sample weighting should be data-dependent and task-dependent. The importance of a sample for the objective function optimization is determined by its uncertainties to both object classification and bounding box regression tasks. To this end, we devise a general loss function to cover most region-based object detectors with various sampling strategies, and then based on it we propose a unified sample weighting network to predict a sample's task weights. Our framework is simple yet effective. It leverages the samples' uncertainty distributions on classification loss, regression loss, IoU, and probability score, to predict sample weights. Our approach has several advantages: (i). It jointly learns sample weights for both classification and regression tasks, which differentiates it from most previous work. (ii). It is a data-driven process, so it avoids some manual parameter tuning. (iii). It can be effortlessly plugged into most object detectors and achieves noticeable performance improvements without affecting their inference time. Our approach has been thoroughly evaluated with recent object detection frameworks and it can consistently boost the detection accuracy. Code has been made available at \url{https://github.com/caiqi/sample-weighting-network}.
We report on the development of machine learning models for classifying C100 superconducting radio-frequency (SRF) cavity faults in the Continuous Electron Beam Accelerator Facility (CEBAF) at Jefferson Lab. CEBAF is a continuous-wave recirculating linac utilizing 418 SRF cavities to accelerate electrons up to 12 GeV through 5-passes. Of these, 96 cavities (12 cryomodules) are designed with a digital low-level RF system configured such that a cavity fault triggers waveform recordings of 17 RF signals for each of the 8 cavities in the cryomodule. Subject matter experts (SME) are able to analyze the collected time-series data and identify which of the eight cavities faulted first and classify the type of fault. This information is used to find trends and strategically deploy mitigations to problematic cryomodules. However manually labeling the data is laborious and time-consuming. By leveraging machine learning, near real-time (rather than post-mortem) identification of the offending cavity and classification of the fault type has been implemented. We discuss performance of the ML models during a recent physics run. Results show the cavity identification and fault classification models have accuracies of 84.9% and 78.2%, respectively.
Adversarial examples have pointed out Deep Neural Networks vulnerability to small local noise. It has been shown that constraining their Lipschitz constant should enhance robustness, but make them harder to learn with classical loss functions. We propose a new framework for binary classification, based on optimal transport, which integrates this Lipschitz constraint as a theoretical requirement. We propose to learn 1-Lipschitz networks using a new loss that is an hinge regularized version of the Kantorovich-Rubinstein dual formulation for the Wasserstein distance estimation. This loss function has a direct interpretation in terms of adversarial robustness together with certifiable robustness bound. We also prove that this hinge regularized version is still the dual formulation of an optimal transportation problem, and has a solution. We also establish several geometrical properties of this optimal solution, and extend the approach to multi-class problems. Experiments show that the proposed approach provides the expected guarantees in terms of robustness without any significant accuracy drop. The adversarial examples, on the proposed models, visibly and meaningfully change the input providing an explanation for the classification.
In this work, we develop efficient disruptions of black-box image translation deepfake generation systems. We are the first to demonstrate black-box deepfake generation disruption by presenting image translation formulations of attacks initially proposed for classification models. Nevertheless, a naive adaptation of classification black-box attacks results in a prohibitive number of queries for image translation systems in the real-world. We present a frustratingly simple yet highly effective algorithm Leaking Universal Perturbations (LUP), that significantly reduces the number of queries needed to attack an image. LUP consists of two phases: (1) a short leaking phase where we attack the network using traditional black-box attacks and gather information on successful attacks on a small dataset and (2) and an exploitation phase where we leverage said information to subsequently attack the network with improved efficiency. Our attack reduces the total number of queries necessary to attack GANimation and StarGAN by 30%.