Parkinson's disease patients develop different speech impairments that affect their communication capabilities. The automatic assessment of the speech of the patients allows the development of computer aided tools to support the diagnosis and the evaluation of the disease severity. This paper introduces a methodology to classify Parkinson's disease from speech in three different languages: Spanish, German, and Czech. The proposed approach considers convolutional neural networks trained with time frequency representations and a transfer learning strategy among the three languages. The transfer learning scheme aims to improve the accuracy of the models when the weights of the neural network are initialized with utterances from a different language than the used for the test set. The results suggest that the proposed strategy improves the accuracy of the models in up to 8\% when the base model used to initialize the weights of the classifier is robust enough. In addition, the results obtained after the transfer learning are in most cases more balanced in terms of specificity-sensitivity than those trained without the transfer learning strategy.

The use of machine learning to model motor learning mechanisms is still limited, while it could help to design novel interactive systems for movement learning or rehabilitation. This approach requires to account for the motor variability induced by motor learning mechanisms. This represents specific challenges concerning fast adaptability of the computational models, from small variations to more drastic changes, including new movement classes. We propose a short review on machine learning based movement models and their existing adaptation mechanisms. We discuss the current challenges for applying these models in motor learning support systems, delineating promising research directions at the intersection of machine learning and motor learning.

Artificial neural networks (ANNs) especially deep convolutional networks are very popular these days and have been proved to successfully offer quite reliable solutions to many vision problems. However, the use of deep neural networks is widely impeded by their intensive computational and memory cost. In this paper, we propose a novel efficient network pruning method that is suitable for both non-structured and structured channel-level pruning. Our proposed method tightens a sparsity constraint by gradually removing network parameters or filter channels based on a criterion and a schedule. The attractive fact that the network size keeps dropping throughout the iterations makes it suitable for the pruning of any untrained or pre-trained network. Because our method uses a L0 constraint instead of the L1 penalty, it does not introduce any bias in the training parameters or filter channels. Furthermore, the L0 constraint makes it easy to directly specify the desired sparsity level during the network pruning process. Finally, experimental validation on synthetic and real datasets both show that the proposed method obtains better or competitive performance compared to other states of art network pruning methods.

Projected Gradient Descent (PGD) based adversarial training has become one of the most prominent methods for building robust deep neural network models. However, the computational complexity associated with this approach, due to the maximization of the loss function when finding adversaries, is a longstanding problem and may be prohibitive when using larger and more complex models. In this paper, we propose a modification of the PGD method for adversarial training and demonstrate that models can be trained much more efficiently without any loss in accuracy on natural and adversarial samples. We argue that the initial phase of adversarial training is redundant and can be replaced with natural training thereby increasing the computational efficiency significantly. We support our argument with insights on the nature of the adversaries and their relative strength during the training process. We show that our proposed method can reduce the training time to up to 38\% of the original training time with comparable model accuracy and generalization on various strengths of adversarial attacks.

The simplified series is obtained by first applying their univariate technique to each of the variables independently; that is, each univariate batch of data is separated into variable-length subsequences, and the obtained subsequences are then clustered as explained in Section 4.1. With this process, a set of representative univariate subsequences is obtained for each variable. Each new multivariate batch of data is then represented by a vector of distances, (d 1,d 2,...,d l), where d j represents the Euclidean distance between the j th variable-length subsequence of the new batch and its corresponding representative subsequence. As with their univariate technique, the reference of normality that is considered by this method is the same time series. The technique proposed by Hu et al. [2019] is also based on reducing the dimensionality of the time series and allows us to detect variable-length discords, while using the same time series as the reference of normality. This is based on the fact that the most unusual subsequences tend to have local regions with significantly different densities (points that are similar) in comparison to the other subsequences in the series. Each point in the new univariate time series describes the density of a local region of the input multivariate time series obtained by a sliding window. This series is also used to obtain the variable-length subsequences. Discords are identified using the Euclidean and Bhattacharyya distances simultaneously.

Existing uncertainty modeling approaches try to detect an out-of-distribution point from the in-distribution dataset. We extend this argument to detect finer-grained uncertainty that distinguishes between (a). certain points, (b). uncertain points but within the data distribution, and (c). out-of-distribution points. Our method corrects overconfident NN decisions, detects outlier points and learns to say ``I don't know'' when uncertain about a critical point between the top two predictions. In addition, we provide a mechanism to quantify class distributions overlap in the decision manifold and investigate its implications in model interpretability. Our method is two-step: in the first step, the proposed method builds a class distribution using Kernel Activation Vectors (kav) extracted from the Network. In the second step, the algorithm determines the confidence of a test point by a hierarchical decision rule based on the chi-squared distribution of squared Mahalanobis distances. Our method sits on top of a given Neural Network, requires a single scan of training data to estimate class distribution statistics, and is highly scalable to deep networks and wider pre-softmax layer. As a positive side effect, our method helps to prevent adversarial attacks without requiring any additional training. It is directly achieved when the Softmax layer is substituted by our robust uncertainty layer at the evaluation phase.

In the big data era, many organizations face the dilemma of data sharing. Regular data sharing is often necessary for human-centered discussion and communication, especially in medical scenarios. However, unprotected data sharing may also lead to data leakage. Inspired by adversarial attack, we propose a method for data encryption, so that for human beings the encrypted data look identical to the original version, but for machine learning methods they are misleading. To show the effectiveness of our method, we collaborate with the Beijing Tiantan Hospital, which has a world leading neurological center. We invite $3$ doctors to manually inspect our encryption method based on real world medical images. The results show that the encrypted images can be used for diagnosis by the doctors, but not by machine learning methods.

Colorectal polyps are abnormalities in the colon tissue that can develop into colorectal cancer. The survival rate for patients is higher when the disease is detected at an early stage and polyps can be removed before they develop into malignant tumors. Deep learning methods have become the state of art in automatic polyp detection. However, the performance of current models heavily relies on the size and quality of the training datasets. Endoscopic video sequences tend to be corrupted by different artifacts affecting visibility and hence, the detection rates. In this work, we analyze the effects that artifacts have in the polyp localization problem. For this, we evaluate the RetinaNet architecture, originally defined for object localization. We also define a model inspired by the learning without forgetting framework, which allows us to employ artifact detection knowledge in the polyp localization problem. Finally, we perform several experiments to analyze the influence of the artifacts in the performance of these models. To our best knowledge, this is the first extensive analysis of the influence of artifact in polyp localization and the first work incorporating learning without forgetting ideas for simultaneous artifact and polyp localization tasks.

Federated learning (FL) is a heavily promoted approach for training ML models on sensitive data, e.g., text typed by users on their smartphones. FL is expressly designed for training on data that are unbalanced and non-iid across the participants. To ensure privacy and integrity of the federated model, latest FL approaches use differential privacy or robust aggregation to limit the influence of "outlier" participants. First, we show that on standard tasks such as next-word prediction, many participants gain no benefit from FL because the federated model is less accurate on their data than the models they can train locally on their own. Second, we show that differential privacy and robust aggregation make this problem worse by further destroying the accuracy of the federated model for many participants. Then, we evaluate three techniques for local adaptation of federated models: fine-tuning, multi-task learning, and knowledge distillation. We analyze where each technique is applicable and demonstrate that all participants benefit from local adaptation. Participants whose local models are poor obtain big accuracy improvements over conventional FL. Participants whose local models are better than the federated model and who have no incentive to participate in FL today improve less, but sufficiently to make the adapted federated model better than their local models.

Recent powerful pre-trained language models have achieved remarkable performance on most of the popular datasets for reading comprehension. It is time to introduce more challenging datasets to push the development of this field towards more comprehensive reasoning of text. In this paper, we introduce a new Reading Comprehension dataset requiring logical reasoning (ReClor) extracted from standardized graduate admission examinations. As earlier studies suggest, human-annotated datasets usually contain biases, which are often exploited by models to achieve high accuracy without truly understanding the text. In order to comprehensively evaluate the logical reasoning ability of models on ReClor, we propose to identify biased data points and separate them into EASY set while the rest as HARD set. Empirical results show that state-of-the-art models have an outstanding ability to capture biases contained in the dataset with high accuracy on EASY set. However, they struggle on HARD set with poor performance near that of random guess, indicating more research is needed to essentially enhance the logical reasoning ability of current models. 1