Deep Bayesian neural networks (BNNs) are a powerful tool, though computationally demanding, to perform parameter estimation while jointly estimating uncertainty around predictions. BNNs are typically implemented using arbitrary normal-distributed prior distributions on the model parameters. Here, we explore the effects of different prior distributions on classification tasks in BNNs and evaluate the evidence supporting the predictions based on posterior probabilities approximated by Markov Chain Monte Carlo sampling and by computing Bayes factors. We show that the choice of priors has a substantial impact on the ability of the model to confidently assign data to the correct class (true positive rates). Prior choice also affects significantly the ability of a BNN to identify out-of-distribution instances as unknown (false positive rates). When comparing our results against neural networks (NN) with Monte Carlo dropout we found that BNNs generally outperform NNs. Finally, in our tests we did not find a single best choice as prior distribution. Instead, each dataset yielded the best results under a different prior, indicating that testing alternative options can improve the performance of BNNs.

Feature selection is one of the most challenging issues in machine learning, especially while working with high dimensional data. In this paper, we address the problem of feature selection and propose a new approach called Evolving Fast and Slow. This new approach is based on using two parallel genetic algorithms having high and low mutation rates, respectively. Evolving Fast and Slow requires a new parallel architecture combining an automatic system that evolves fast and an effortful system that evolves slow. With this architecture, exploration and exploitation can be done simultaneously and in unison. Evolving fast, with high mutation rate, can be useful to explore new unknown places in the search space with long jumps; and Evolving Slow, with low mutation rate, can be useful to exploit previously known places in the search space with short movements. Our experiments show that Evolving Fast and Slow achieves very good results in terms of both accuracy and feature elimination.

Indoor localization has become an important issue for wireless sensor networks. This paper presents a zoning-based localization technique that uses WiFi signals and works efficiently in indoor environments. The targeted area is composed of several zones, the objective being to determine the zone of the sensor using an observation model based on statistical learning.

Machine Learning for graphs is nowadays a research topic of consolidated relevance. Common approaches in the field typically resort to complex deep neural network architectures and demanding training algorithms, highlighting the need for more efficient solutions. The class of Reservoir Computing (RC) models can play an important role in this context, enabling to develop fruitful graph embeddings through untrained recursive architectures. In this paper, we study progressive simplifications to the design strategy of RC neural networks for graphs. Our core proposal is based on shaping the organization of the hidden neurons to follow a ring topology. Experimental results on graph classification tasks indicate that ring-reservoirs architectures enable particularly effective network configurations, showing consistent advantages in terms of predictive performance.

Brain waves vary between people. An obvious way to improve automatic sleep staging for longitudinal sleep monitoring is personalization of algorithms based on individual characteristics extracted from the first night of data. As a single night is a very small amount of data to train a sleep staging model, we propose a Kullback-Leibler (KL) divergence regularized transfer learning approach to address this problem. We employ the pretrained SeqSleepNet (i.e. the subject independent model) as a starting point and finetune it with the single-night personalization data to derive the personalized model. This is done by adding the KL divergence between the output of the subject independent model and the output of the personalized model to the loss function during finetuning. In effect, KL-divergence regularization prevents the personalized model from overfitting to the single-night data and straying too far away from the subject independent model. Experimental results on the Sleep-EDF Expanded database with 75 subjects show that sleep staging personalization with a single-night data is possible with help of the proposed KL-divergence regularization. On average, we achieve a personalized sleep staging accuracy of 79.6%, a Cohen's kappa of 0.706, a macro F1-score of 73.0%, a sensitivity of 71.8%, and a specificity of 94.2%. We find both that the approach is robust against overfitting and that it improves the accuracy by 4.5 percentage points compared to non-personalization and 2.2 percentage points compared to personalization without regularization.

Heavy rainfall events can cause major disruption to human activities. It is desirable to predict these events ahead of time so that decision makers can take action to protect life, property and prosperity. Nowcasting, or short-term forecasting from observations, remains an important tool in predicting these events. The essential goals of nowcasting are identical to those of all weather forecasting, with the only difference being the spatial and temporal scales involved. The World Meteorological Organization (WMO, 2016) distinguishes among the various forecasting time horizons as: "Usually forecasts for the next 0-2 hours are called nowcasting, from 2-12 hours very short-range forecasting (VSRF), and short-range forecasting beyond that; but the capabilities of the different ranges can vary upon variables and weather situations." Radar-based nowcasting emerged in an era of mainly synoptic and mesoscale weather prediction. Predicting rainfall during that time was a challenge for numerical weather prediction (NWP) models, since computational restrictions limited the resolution at which NWP models could operate. As a result, NWP models were able to capture mesoscale weather patterns such as fronts, but not the smaller-scale convective patterns that occur within mesoscale systems. Thus, these models had limited utility in predicting rainfall in the early hours of the forecast because of its dependence on the unrepresented small scales.

In many scientific problems such as video surveillance, modern genomic analysis, and clinical studies, data are often collected from diverse domains across time that exhibit time-dependent heterogeneous properties. It is important to not only integrate data from multiple sources (called multiview data), but also to incorporate time dependency for deep understanding of the underlying system. Latent factor models are popular tools for exploring multi-view data. However, it is frequently observed that these models do not perform well for complex systems and they are not applicable to time-series data. Therefore, we propose a generative model based on variational autoencoder and recurrent neural network to infer the latent dynamic factors for multivariate timeseries data. This approach allows us to identify the disentangled latent embeddings across multiple modalities while accounting for the time factor. We invoke our proposed model for analyzing three datasets on which we demonstrate the effectiveness and the interpretability of the model.

Federated learning is a rapidly evolving application of distributed optimization for estimation and learning problems in large-scale networks of remote clients [13]. These systems present new challenges, as they are characterized by heterogeneity in computational resources and data across the network, unreliable communication, massive scale, and privacy constraints [16]. A typical application is for developers of cell phones and cellular applications to model the usage of software and devices across millions or even billions of users. Distributed optimization has a rich history and extensive literature (e.g., see the sources [2, 5, 8, 31, 15, 24] and references therein), and federated learning has led to a flurry of interest in the area. A number of different procedures have been proposed for federated learning and related problems, using methods based on stochastic gradient methods or proximal procedures. Notably, McMahan et al. [18] introduced the FedSGD and FedAvg algorithms, which both adapt the classical stochastic gradient method to the federated setting, considering the possibility that clients may fail and may only be subsampled on each round of computation. Another recent proposal has been to use regularized local problems to mitigate possible issues that arise with device heterogeneity and failures [17]. These authors propose the FedProx procedure, an algorithm that applied averaged proximal updates to solve federated minimization problems. Currently, the convergence theory and correctness of these methods is currently lacking, and practitioners have documented failures of convergence in certain settings (e.g., see Figure 3 and related discussion in the work [18]).

We publish dataset of visual P300 BCI performed in Virtual Reality (VR) game Raccoons versus Demons (RvD). Data contains reach labels incorporating information about stimulus chosen enabling us to estimate model's confidence at each stimulus prediction stage. Data and experiments code are available at https://gitlab.com/impulse-neiry_public/raccoons-vs-demons

This paper presents channel-aware adversarial attacks against deep learning-based wireless signal classifiers. There is a transmitter that transmits signals with different modulation types. A deep neural network is used at each receiver to classify its over-the-air received signals to modulation types. In the meantime, an adversary transmits an adversarial perturbation (subject to a power budget) to fool receivers into making errors in classifying signals that are received as superpositions of transmitted signals and adversarial perturbations. First, these evasion attacks are shown to fail when channels are not considered in designing adversarial perturbations. Then realistic attacks are presented by considering channel effects from the adversary to each receiver. After showing that a channel-aware attack is selective (i.e., it affects only the receiver whose channel is considered in the perturbation design), a broadcast adversarial attack is presented by crafting a common adversarial perturbation to simultaneously fool classifiers at different receivers. The major vulnerability of modulation classifiers to over-the-air adversarial attacks is shown by accounting for different levels of information available about channel, transmitter input, and classifier model. Finally, a certified defense based on randomized smoothing that augments training data with noise is introduced to make modulation classifier robust to adversarial perturbations.