Deep learning models for MRI classification face two recurring problems: they are typically limited by low sample size, and are abstracted by their own complexity (the "black box problem"). In this paper, we train a convolutional neural network (CNN) with the largest multi-source, functional MRI (fMRI) connectomic dataset ever compiled, consisting of 43,858 datapoints. We apply this model to a cross-sectional comparison of autism (ASD) vs typically developing (TD) controls that has proved difficult to characterise with inferential statistics. To contextualise these findings, we additionally perform classifications of gender and task vs rest. Employing class-balancing to build a training set, we trained 3$\times$300 modified CNNs in an ensemble model to classify fMRI connectivity matrices with overall AUROCs of 0.6774, 0.7680, and 0.9222 for ASD vs TD, gender, and task vs rest, respectively. Additionally, we aim to address the black box problem in this context using two visualization methods. First, class activation maps show which functional connections of the brain our models focus on when performing classification. Second, by analyzing maximal activations of the hidden layers, we were also able to explore how the model organizes a large and mixed-centre dataset, finding that it dedicates specific areas of its hidden layers to processing different covariates of data (depending on the independent variable analyzed), and other areas to mix data from different sources. Our study finds that deep learning models that distinguish ASD from TD controls focus broadly on temporal and cerebellar connections, with a particularly high focus on the right caudate nucleus and paracentral sulcus.

The design of symbol detectors in digital communication systems has traditionally relied on statistical channel models that describe the relation between the transmitted symbols and the observed signal at the receiver. Here we review a data-driven framework to symbol detection design which combines machine learning (ML) and model-based algorithms. In this hybrid approach, well-known channelmodel-based algorithms such as the Viterbi method, BCJR detection, and multiple-input multiple-output (MIMO) soft interference cancellation (SIC) are augmented with MLbased algorithms to remove their channel-model-dependence, allowing the receiver to learn to implement these algorithms solely from data. The resulting data-driven receivers are most suitable for systems where the underlying channel models are poorly understood, highly complex, or do not well-capture the underlying physics. Our approach is unique in that it only replaces the channel-model-based computations with dedicated neural networks that can be trained from a small amount of data, while keeping the general algorithm intact. Our results demonstrate that these techniques can yield near-optimal performance of model-based algorithms without knowing the exact channel input-output statistical relationship and in the presence of channel state information uncertainty. I. INTRODUCTION In digital communication systems, the receiver is required to reliably recover the transmitted symbols from the observed channel output. This task is commonly referred to as symbol detection. Conventional symbol detection algorithms, such as those based on the maximum a-posteriori probability (MAP) rule, require complete knowledge of the underlying channel model and its parameters [1], [2]. This work was supported in part by the US - Israel Binational Science Foundation under grant No. 2026094, by the Israel Science Foundation under grant No. 0100101, and by the Office of the Naval Research under grant No. 18-1-2191. N. Shlezinger and Y. C. Eldar are with the Faculty of Math and CS, Weizmann Institute of Science, Rehovot, Israel (email: nirshlezinger1@gmail.com; yonina@weizmann.ac.il). Furthermore, when the channel models are known, many detection algorithms rely on channel state information (CSI), i.e., the instantaneous parameters of the channel model, for detection. Therefore, conventional channel-model-based techniques require the instantaneous CSI to be estimated.

Single-cell RNA sequencing (scRNA-seq) has revolutionized biological discovery, providing an unbiased picture of cellular heterogeneity in tissues. While scRNA-seq has been used extensively to provide insight into health and disease, it has not been used for disease prediction or diagnostics. Graph Attention Networks have proven to be versatile for a wide range of tasks by learning from both original features and graph structures. Here we present a graph attention model for predicting disease state from single-cell data on a large dataset of Multiple Sclerosis (MS) patients. MS is a disease of the central nervous system that is difficult to diagnose. We train our model on single-cell data obtained from blood and cerebrospinal fluid (CSF) for a cohort of seven MS patients and six healthy adults (HA), resulting in 66,667 individual cells. We achieve $\mathbf{92}$ \% accuracy in predicting MS, outperforming other state-of-the-art methods such as a graph convolutional network, random forest, and multi-layer perceptron. Further, we use the learned graph attention model to get insight into the features (cell types and genes) that are important for this prediction. The graph attention model also allow us to infer a new feature space for the cells that emphasizes the difference between the two conditions. Finally we use the attention weights to learn a new low-dimensional embedding which we visualize with PHATE and UMAP. To the best of our knowledge, this is the first effort to use graph attention, and deep learning in general, to predict disease state from single-cell data. We envision applying this method to single-cell data for other diseases.

Deep residual networks (ResNets) have demonstrated better generalization performance than deep feedforward networks (FFNets). However, the theory behind such a phenomenon is still largely unknown. This paper studies this fundamental problem in deep learning from a so-called "neural tangent kernel" perspective. Specifically, we first show that under proper conditions, as the width goes to infinity, training deep ResNets can be viewed as learning reproducing kernel functions with some kernel function. We then compare the kernel of deep ResNets with that of deep FFNets and discover that the class of functions induced by the kernel of FFNets is asymptotically not learnable, as the depth goes to infinity. In contrast, the class of functions induced by the kernel of ResNets does not exhibit such degeneracy. Our discovery partially justifies the advantages of deep ResNets over deep FFNets in generalization abilities. Numerical results are provided to support our claim.

In this work we propose a novel self-attention mechanism model to address electricity theft detection on an imbalanced realistic dataset that presents a daily electricity consumption provided by State Grid Corporation of China. Our key contribution is the introduction of a multi-head self-attention mechanism concatenated with dilated convolutions and unified by a convolution of kernel size $1$. Moreover, we introduce a binary input channel (Binary Mask) to identify the position of the missing values, allowing the network to learn how to deal with these values. Our model achieves an AUC of $0.926$ which is an improvement in more than $17\%$ with respect to previous baseline work. The code is available on GitHub at https://github.com/neuralmind-ai/electricity-theft-detection-with-self-attention.

We propose a mechanism for distributed radio resource management using multi-agent deep reinforcement learning (RL) for interference mitigation in wireless networks. We equip each transmitter in the network with a deep RL agent, which receives partial delayed observations from its associated users, while also exchanging observations with its neighboring agents, and decides on which user to serve and what transmit power to use at each scheduling interval. Our proposed framework enables the agents to make decisions simultaneously and in a distributed manner, without any knowledge about the concurrent decisions of other agents. Moreover, our design of the agents' observation and action spaces is scalable, in the sense that an agent trained on a scenario with a specific number of transmitters and receivers can be readily applied to scenarios with different numbers of transmitters and/or receivers. Simulation results demonstrate the superiority of our proposed approach compared to decentralized baselines in terms of the tradeoff between average and 5 th percentile user rates, while achieving performance close to, and even in certain cases outperforming, that of a centralized information-theoretic scheduling algorithm. We also show that our trained agents are robust and maintain their performance gains when experiencing mismatches between training and testing deployments. I. INTRODUCTION One of the key drivers for improving throughput in future wireless networks, including fifth generation mobile networks (5G), is the densification achieved by deploying more base stations. The authors are with Intel Corporation, Santa Clara, CA 95054. The rise of such ultra-dense network paradigms implies that the limited physical wireless resources (in time, frequency, etc.) need to support an increasing number of simultaneous transmissions. Effective radio resource management procedures are, therefore, critical to mitigate the interference among such concurrent transmissions and achieve the desired performance enhancement in these ultra-dense environments. The radio resource management problem is in general non-convex and therefore computationally complex, especially as the network size increases. There is a rich literature of centralized and distributed algorithms for radio resource management, using various techniques in different areas such as geometric programming [1], weighted minimum mean square optimization [2], game theory [3], information theory [4], [5], and fractional programming [6].

We address two fundamental questions about graph neural networks (GNNs). First, we prove that several important graph properties cannot be computed by GNNs that rely entirely on local information. Such GNNs include the standard message passing models, and more powerful spatial variants that exploit local graph structure (e.g., via relative orientation of messages, or local port ordering) to distinguish neighbors of each node. Our treatment includes a novel graph-theoretic formalism. Second, we provide the first data dependent generalization bounds for message passing GNNs. This analysis explicitly accounts for the local permutation invariance of GNNs. Our bounds are much tighter than existing VC-dimension based guarantees for GNNs, and are comparable to Rademacher bounds for recurrent neural networks.

Time series forecasting is often fundamental to scientific and engineering problems and enables decision making. With ever increasing data set sizes, a trivial solution to scale up predictions is to assume independence between interacting time series. However, modeling statistical dependencies can improve accuracy and enable analysis of interaction effects. Deep learning methods are well suited for this problem, but multi-variate models often assume a simple parametric distribution and do not scale to high dimensions. In this work we model the multi-variate temporal dynamics of time series via an autoregressive deep learning model, where the data distribution is represented by a conditioned normalizing flow. This combination retains the power of autoregressive models, such as good performance in extrapolation into the future, with the flexibility of flows as a general purpose high-dimensional distribution model, while remaining computationally tractable. We show that it improves over the state-of-the-art for standard metrics on many real-world data sets with several thousand interacting time-series.

We derive an unbiased estimator for expectations over discrete random variables based on sampling without replacement, which reduces variance as it avoids duplicate samples. We show that our estimator can be derived as the Rao-Blackwellization of three different estimators. Combining our estimator with REINFORCE, we obtain a policy gradient estimator and we reduce its variance using a built-in control variate which is obtained without additional model evaluations. The resulting estimator is closely related to other gradient estimators. Experiments with a toy problem, a categorical Variational Auto-Encoder and a structured prediction problem show that our estimator is the only estimator that is consistently among the best estimators in both high and low entropy settings.

Speaker recognition systems based on deep speaker embeddings have achieved significant performance in controlled conditions according to the results obtained for early NIST SRE (Speaker Recognition Evaluation) datasets. From the practical point of view, taking into account the increased interest in virtual assistants (such as Amazon Alexa, Google Home, AppleSiri, etc.), speaker verification on short utterances in uncontrolled noisy environment conditions is one of the most challenging and highly demanded tasks. This paper presents approaches aimed to achieve two goals: a) improve the quality of far-field speaker verification systems in the presence of environmental noise, reverberation and b) reduce the system qualitydegradation for short utterances. For these purposes, we considered deep neural network architectures based on TDNN (TimeDelay Neural Network) and ResNet (Residual Neural Network) blocks. We experimented with state-of-the-art embedding extractors and their training procedures. Obtained results confirm that ResNet architectures outperform the standard x-vector approach in terms of speaker verification quality for both long-duration and short-duration utterances. We also investigate the impact of speech activity detector, different scoring models, adaptation and score normalization techniques. The experimental results are presented for publicly available data and verification protocols for the VoxCeleb1, VoxCeleb2, and VOiCES datasets.