Improving subjective sound quality of enhanced signals is one of the most important missions in speech enhancement. For evaluating the subjective quality, several methods related to perceptually-motivated objective sound quality assessment (OSQA) have been proposed such as PESQ (perceptual evaluation of speech quality). However, direct use of such measures for training deep neural network (DNN) is not allowed in most cases because popular OSQAs are non-differentiable with respect to DNN parameters. Therefore, the previous study has proposed to approximate the score of OSQAs by an auxiliary DNN so that its gradient can be used for training the primary DNN. One problem with this approach is instability of the training caused by the approximation error of the score. To overcome this problem, we propose to use stabilization techniques borrowed from reinforcement learning. The experiments, aimed to increase the score of PESQ as an example, show that the proposed method (i) can stably train a DNN to increase PESQ, (ii) achieved the state-of-the-art PESQ score on a public dataset, and (iii) resulted in better sound quality than conventional methods based on subjective evaluation.

This paper investigates a self-adaptation method for speech enhancement using auxiliary speaker-aware features; we extract a speaker representation used for adaptation directly from the test utterance. Conventional studies of deep neural network (DNN)--based speech enhancement mainly focus on building a speaker independent model. Meanwhile, in speech applications including speech recognition and synthesis, it is known that model adaptation to the target speaker improves the accuracy. Our research question is whether a DNN for speech enhancement can be adopted to unknown speakers without any auxiliary guidance signal in test-phase. To achieve this, we adopt multi-task learning of speech enhancement and speaker identification, and use the output of the final hidden layer of speaker identification branch as an auxiliary feature. In addition, we use multi-head self-attention for capturing long-term dependencies in the speech and noise. Experimental results on a public dataset show that our strategy achieves the state-of-the-art performance and also outperform conventional methods in terms of subjective quality.

A well-known issue of Batch Normalization is its significantly reduced effectiveness in the case of small mini-batch sizes. When a mini-batch contains few examples, the statistics upon which the normalization is defined cannot be reliably estimated from it during a training iteration. To address this problem, we present Cross-Iteration Batch Normalization (CBN), in which examples from multiple recent iterations are jointly utilized to enhance estimation quality. A challenge of computing statistics over multiple iterations is that the network activations from different iterations are not comparable to each other due to changes in network weights. We thus compensate for the network weight changes via a proposed technique based on Taylor polynomials, so that the statistics can be accurately estimated and batch normalization can be effectively applied. On object detection and image classification with small mini-batch sizes, CBN is found to outperform the original batch normalization and a direct calculation of statistics over previous iterations without the proposed compensation technique.

We present a new method for evaluating and training unnormalized density models. Our approach only requires access to the gradient of the unnormalized model's log-density. We estimate the Stein discrepancy between the data density p(x) and the model density q(x) defined by a vector function of the data. We parameterize this function with a neural network and fit its parameters to maximize the discrepancy. This yields a novel goodness-of-fit test which outperforms existing methods on high dimensional data. Furthermore, optimizing $q(x)$ to minimize this discrepancy produces a novel method for training unnormalized models which scales more gracefully than existing methods. The ability to both learn and compare models is a unique feature of the proposed method.

We propose a new optimization formulation for training federated learning models. The standard formulation has the form of an empirical risk minimization problem constructed to find a single global model trained from the private data stored across all participating devices. In contrast, our formulation seeks an explicit trade-off between this traditional global model and the local models, which can be learned by each device from its own private data without any communication. Further, we develop several efficient variants of SGD (with and without partial participation and with and without variance reduction) for solving the new formulation and prove communication complexity guarantees. Notably, our methods are similar but not identical to federated averaging / local SGD, thus shedding some light on the essence of the elusive method. In particular, our methods do not perform full averaging steps and instead merely take steps towards averaging. We argue for the benefits of this new paradigm for federated learning.

Machine learning systems have received much attention recently for their ability to achieve expert-level performance on clinical tasks, particularly in medical imaging. Here, we examine the extent to which state-of-the-art deep learning classifiers trained to yield diagnostic labels from X-ray images are biased with respect to protected attributes. We train convolution neural networks to predict 14 diagnostic labels in three prominent public chest X-ray datasets: MIMIC-CXR, Chest-Xray8, and CheXpert. We then evaluate the TPR disparity - the difference in true positive rates (TPR) and - underdiagnosis rate - the false positive rate of a non-diagnosis - among different protected attributes such as patient sex, age, race, and insurance type. We demonstrate that TPR disparities exist in the state-of-the-art classifiers in all datasets, for all clinical tasks, and all subgroups. We find that TPR disparities are most commonly not significantly correlated with a subgroup's proportional disease burden; further, we find that some subgroups and subsection of the population are chronically underdiagnosed. Such performance disparities have real consequences as models move from papers to products, and should be carefully audited prior to deployment.

Recent advances in the interconnectedness and digitization of industrial machines, known as Industry 4.0, pave the way for new analytical techniques. Indeed, the availability and the richness of production-related data enables new data-driven methods. In this paper, we propose a process mining approach augmented with artificial intelligence that (1) reconstructs the historical workload of a company and (2) predicts the workload using neural networks. Our method relies on logs, representing the history of business processes related to manufacturing. These logs are used to quantify the supply and demand and are fed into a recurrent neural network model to predict customer orders. The corresponding activities to fulfill these orders are then sampled from history with a replay mechanism, based on criteria such as trace frequency and activities similarity. An evaluation and illustration of the method is performed on the administrative processes of Heraeus Materials SA. The workload prediction on a one-year test set achieves an MAPE score of 19% for a one-week forecast. The case study suggests a reasonable accuracy and confirms that a good understanding of the historical workload combined to articulated predictions are of great help for supporting management decisions and can decrease costs with better resources planning on a medium-term level.

Recently, the problem of allocating one resource per agent with initial endowments (house markets) has seen a renewed interest: indeed, while in the domain of strict preferences the Top Trading Cycle algorithm is known to be the only procedure guaranteeing Pareto-optimality, individual rationality, and strategy proofness. However, the situation differs in the single-peaked domain. Indeed, Bade presented the Crawler, an alternative procedure enjoying the same properties, with the additional advantage of being implementable in obviously dominant strategies. In this paper we further investigate the Crawler and propose the Diver, a variant which checks optimally whether an allocation is Pareto-optimal for single-peaked preferences, thus improving over known techniques used for checking Pareto-optimality in more general domains. We also prove that the Diver is asymptotically optimal in terms of communication complexity.

Student dropout prediction provides an opportunity to improve student engagement, which maximizes the overall effectiveness of learning experiences. However, researches on student dropout were mainly conducted on school dropout or course dropout, and study session dropout in a mobile learning environment has not been considered thoroughly. In this paper, we investigate the study session dropout prediction problem in a mobile learning environment. First, we define the concept of the study session, study session dropout and study session dropout prediction task in a mobile learning environment. Based on the definitions, we propose a novel Transformer based model for predicting study session dropout, DAS: Deep Attentive Study Session Dropout Prediction in Mobile Learning Environment. DAS has an encoder-decoder structure which is composed of stacked multi-head attention and point-wise feed-forward networks. The deep attentive computations in DAS are capable of capturing complex relations among dynamic student interactions. To the best of our knowledge, this is the first attempt to investigate study session dropout in a mobile learning environment. Empirical evaluations on a large-scale dataset show that DAS achieves the best performance with a significant improvement in area under the receiver operating characteristic curve compared to baseline models.

Knowledge tracing, the act of modeling a student's knowledge through learning activities, is an extensively studied problem in the field of computer-aided education. Although models with attention mechanism have outperformed traditional approaches such as Bayesian knowledge tracing and collaborative filtering, they share two limitations. Firstly, the models rely on shallow attention layers and fail to capture complex relations among exercises and responses over time. Secondly, different combinations of queries, keys and values for the self-attention layer for knowledge tracing were not extensively explored. Usual practice of using exercises and interactions (exercise-response pairs) as queries and keys/values respectively lacks empirical support. In this paper, we propose a novel Transformer based model for knowledge tracing, SAINT: Separated Self-AttentIve Neural Knowledge Tracing. SAINT has an encoder-decoder structure where exercise and response embedding sequence separately enter the encoder and the decoder respectively, which allows to stack attention layers multiple times. To the best of our knowledge, this is the first work to suggest an encoder-decoder model for knowledge tracing that applies deep self-attentive layers to exercises and responses separately. The empirical evaluations on a large-scale knowledge tracing dataset show that SAINT achieves the state-of-the-art performance in knowledge tracing with the improvement of AUC by 1.8% compared to the current state-of-the-art models.