Deep neural networks (DNNs) for supervised learning can be viewed as a pipeline of the feature extractor (i.e., last hidden layer) and a linear classifier (i.e., output layer) that are trained jointly with stochastic gradient descent (SGD) on the loss function (e.g., cross-entropy). In each epoch, the true gradient of the loss function is estimated using a mini-batch sampled from the training set and model parameters are then updated with the mini-batch gradients. Although the latter provides an unbiased estimation of the former, they are subject to substantial variances derived from the size and number of sampled mini-batches, leading to noisy and jumpy updates. To stabilize such undesirable variance in estimating the true gradients, we propose In-Training Representation Alignment (ITRA) that explicitly aligns feature distributions of two different mini-batches with a matching loss in the SGD training process. We also provide a rigorous analysis of the desirable effects of the matching loss on feature representation learning: (1) extracting compact feature representation; (2) reducing over-adaption on mini-batches via an adaptively weighting mechanism; and (3) accommodating to multi-modalities. Finally, we conduct large-scale experiments on both image and text classifications to demonstrate its superior performance to the strong baselines.

There are increasing demands for understanding deep neural networks' (DNNs) behavior spurred by growing security and/or transparency concerns. Due to multi-layer nonlinearity of the deep neural network architectures, explaining DNN predictions still remains as an open problem, preventing us from gaining a deeper understanding of the mechanisms. To enhance the explainability of DNNs, we estimate the input feature's attributions to the prediction task using divergence and flux. Inspired by the divergence theorem in vector analysis, we develop a novel Negative Flux Aggregation (NeFLAG) formulation and an efficient approximation algorithm to estimate attribution map. Unlike the previous techniques, ours doesn't rely on fitting a surrogate model nor need any path integration of gradients. Both qualitative and quantitative experiments demonstrate a superior performance of NeFLAG in generating more faithful attribution maps than the competing methods. Our code is available at https://github.com/xinli0928/NeFLAG.

Transformers have improved the state-of-the-art in various natural language processing and computer vision tasks. However, the success of the Transformer model has not yet been duly explained. Current explanation techniques, which dissect either the self-attention mechanism or gradient-based attribution, do not necessarily provide a faithful explanation of the inner workings of Transformers due to the following reasons: first, attention weights alone without considering the magnitudes of feature values are not adequate to reveal the self-attention mechanism; second, whereas most Transformer explanation techniques utilize self-attention module, the skip-connection module, contributing a significant portion of information flows in Transformers, has not yet been sufficiently exploited in explanation; third, the gradient-based attribution of individual feature does not incorporate interaction among features in explaining the model's output. In order to tackle the above problems, we propose a novel Transformer explanation technique via attentive class activation tokens, aka, AttCAT, leveraging encoded features, their gradients, and their attention weights to generate a faithful and confident explanation for Transformer's output. Extensive experiments are conducted to demonstrate the superior performance of AttCAT, which generalizes well to different Transformer architectures, evaluation metrics, datasets, and tasks, to the baseline methods. Our code is available at: https://github.com/qiangyao1988/AttCAT.

Convolutional neural networks (CNNs) have achieved state-of-the-art performance on various tasks in computer vision. However, recent studies demonstrate that these models are vulnerable to carefully crafted adversarial samples and suffer from a significant performance drop when predicting them. Many methods have been proposed to improve adversarial robustness (e.g., adversarial training and new loss functions to learn adversarially robust feature representations). Here we offer a unique insight into the predictive behavior of CNNs that they tend to misclassify adversarial samples into the most probable false classes. This inspires us to propose a new Probabilistically Compact (PC) loss with logit constraints which can be used as a drop-in replacement for cross-entropy (CE) loss to improve CNN's adversarial robustness. Specifically, PC loss enlarges the probability gaps between true class and false classes meanwhile the logit constraints prevent the gaps from being melted by a small perturbation. We extensively compare our method with the state-of-the-art using large scale datasets under both white-box and black-box attacks to demonstrate its effectiveness. The source codes are available at https://github.com/xinli0928/PC-LC.

Deep neural networks (DNNs) have became one of the most high performing tools in a broad range of machine learning areas. However, the multilayer non-linearity of the network architectures prevent us from gaining a better understanding of the models’ predictions. Gradient based attribution methods (e.g., Integrated Gradient (IG)) that decipher input features’ contribution to the prediction task have been shown to be highly effective yet requiring a reference input as the anchor for explaining model’s output. The performance of DNN model interpretation can be quite inconsistent with regard to the choice of references. Here we propose an Adversarial Gradient Integration (AGI) method that integrates the gradients from adversarial examples to the target example along the curve of steepest ascent to calculate the resulting contributions from all input features. Our method doesn’t rely on the choice of references, hence can avoid the ambiguity and inconsistency sourced from the reference selection. We demonstrate the performance of our AGI method and compare with competing methods in explaining image classification results. Code is available from https://github.com/pd90506/AGI.

Deep convolutional neural networks (CNNs) trained with logistic and softmax losses have made significant advancement in visual recognition tasks in computer vision. When training data exhibit class imbalances, the class-wise reweighted version of logistic and softmax losses are often used to boost performance of the unweighted version. In this paper, motivated to explain the reweighting mechanism, we explicate the learning property of those two loss functions by analyzing the necessary condition (e.g., gradient equals to zero) after training CNNs to converge to a local minimum. The analysis immediately provides us explanations for understanding (1) quantitative effects of the class-wise reweighting mechanism: deterministic effectiveness for binary classification using logistic loss yet indeterministic for multi-class classification using softmax loss; (2) disadvantage of logistic loss for single-label multi-class classification via one-vs.-all approach, which is due to the averaging effect on predicted probabilities for the negative class (e.g., non-target classes) in the learning process. With the disadvantage and advantage of logistic loss disentangled, we thereafter propose a novel reweighted logistic loss for multi-class classification. Our simple yet effective formulation improves ordinary logistic loss by focusing on learning hard non-target classes (target vs. non-target class in one-vs.-all) and turned out to be competitive with softmax loss. We evaluate our method on several benchmark datasets to demonstrate its effectiveness.

During the COVID-19 pandemic, there has been an emerging need for rapid, dedicated, and point-of-care COVID19 patient disposition techniques to optimize resource utilization and clinical workflow. In view of this need, we present COVIDMobileXpert: a lightweight deep neural network (DNN) based mobile app that can use chest X-ray (CXR) for COVID-19 case screening and radiological trajectory prediction. We design and implement a novel three-player knowledge transfer and distillation (KTD) framework including a pre-trained attending physician (AP) network that extracts CXR imaging features from a large scale of lung disease CXR images, a fine-tuned resident fellow (RF) network that learns the essential CXR imaging features to discriminate COVID-19 from pneumonia and/or normal cases with a small amount of COVID-19 cases, and a trained lightweight medical student (MS) network to perform on-device COVID-19 patient triage and follow-up. To tackle the challenge of vastly similar and dominant fore-and background in medical images, we employ novel loss functions and training schemes for the MS network to learn the robust features. We demonstrate the significant potential of COVID-MobileXpert for rapid deployment via extensive experiments with diverse MS architecture and tuning parameter settings. The source codes for cloud and mobile based implementations are available from the following url: https://github.com/xinli0928/COVID-Xray.

Latent factor collaborative filtering (CF) has been a widely used technique for recommender system by learning the semantic representations of users and items. Recently, explainable recommendation has attracted much attention from research community. However, trade-off exists between explainability and performance of the recommendation where metadata is often needed to alleviate the dilemma. We present a novel feature mapping approach that maps the uninterpretable general features onto the interpretable aspect features, achieving both satisfactory accuracy and explainability in the recommendations by simultaneous minimization of rating prediction loss and interpretation loss. To evaluate the explainability, we propose two new evaluation metrics specifically designed for aspect-level explanation using surrogate ground truth. Experimental results demonstrate a strong performance in both recommendation and explaining explanation, eliminating the need for metadata. Code is available from https://github.com/pd90506/AMCF.

Existing aspect based sentiment analysis (ABSA) approaches leverage various neural network models to extract the aspect sentiments via learning aspect-specific feature representations. However, these approaches heavily rely on manual tagging of user reviews according to the predefined aspects as the input, a laborious and time-consuming process. Moreover, the underlying methods do not explain how and why the opposing aspect level polarities in a user review lead to the overall polarity. In this paper, we tackle these two problems by designing and implementing a new Multiple-Attention Network (MAN) approach for more powerful ABSA without the need for aspect tags using two new tag-free data sets crawled directly from TripAdvisor (https://www.tripadvisor.com). With the Self- and Position-Aware attention mechanism, MAN is capable of extracting both aspect level and overall sentiments from the text reviews using the aspect level and overall customer ratings, and it can also detect the vital aspect(s) leading to the overall sentiment polarity among different aspects via a new aspect ranking scheme. We carry out extensive experiments to demonstrate the strong performance of MAN compared to other state-of-the-art ABSA approaches and the explainability of our approach by visualizing and interpreting attention weights in case studies.