Recent successes in machine learning have led to a shift in the design of autonomous systems, improving performance on existing tasks and rendering new applications possible. Data-focused approaches gain relevance across diverse, intricate applications when developing data collection and curation pipelines becomes more effective than manual behaviour design. The following work aims at increasing the efficiency of this pipeline in two principal ways: by utilising more powerful sources of informative data and by extracting additional information from existing data. In particular, we target three orthogonal fronts: imitation learning, domain adaptation, and transfer from simulation.

The goal of this paper is to study why stochastic gradient descent (SGD) is efficient for neural networks, and how neural net design affects SGD. In particular, we investigate how overparameterization -- an increase in the number of parameters beyond the number of training data -- affects the dynamics of SGD. We introduce a simple concept called gradient confusion. When confusion is high, stochastic gradients produced by different data samples may be negatively correlated, slowing down convergence. But when gradient confusion is low, we show that SGD has better convergence properties than predicted by classical theory. Using theoretical and experimental results, we study how overparameterization affects gradient confusion, and thus the convergence of SGD, on linear models and neural networks. We show that increasing the number of parameters of linear models or increasing the width of neural networks leads to lower gradient confusion, and thus faster and easier model training. We also show how overparameterization by increasing the depth of neural networks results in higher gradient confusion, making deeper models harder to train. Finally, we observe empirically that techniques like batch normalization and skip connections reduce gradient confusion, which helps reduce the training burden of deep networks.

Deep Neural Networks (DNNs) have tremendous potential in advancing the vision for self-driving cars. However, the security of DNN models in this context leads to major safety implications and needs to be better understood. We consider the case study of steering angle prediction from camera images, using the dataset from the 2014 Udacity challenge. We demonstrate for the first time adversarial testing-time attacks for this application for both classification and regression settings. We show that minor modifications to the camera image (an L2 distance of 0.82 for one of the considered models) result in mis-classification of an image to any class of attacker's choice. Furthermore, our regression attack results in a significant increase in Mean Square Error (MSE) by a factor of 69 in the worst case.

Recurrent neural network grammars (RNNG) are generative models of language which jointly model syntax and surface structure by incrementally generating a syntax tree and sentence in a top-down, left-to-right order. Supervised RNNGs achieve strong language modeling and parsing performance, but require an annotated corpus of parse trees. In this work, we experiment with unsupervised learning of RNNGs. Since directly marginalizing over the space of latent trees is intractable, we instead apply amortized variational inference. To maximize the evidence lower bound, we develop an inference network parameterized as a neural CRF constituency parser. On language modeling, unsupervised RNNGs perform as well their supervised counterparts on benchmarks in English and Chinese. On constituency grammar induction, they are competitive with recent neural language models that induce tree structures from words through attention mechanisms.

Computer vision has achieved impressive progress in recent years. Meanwhile, mobile phones have become the primary computing platforms for millions of people. In addition to mobile phones, many autonomous systems rely on visual data for making decisions and some of these systems have limited energy (such as unmanned aerial vehicles also called drones and mobile robots). These systems rely on batteries and energy efficiency is critical. This article serves two main purposes: (1) Examine the state-of-the-art for low-power solutions to detect objects in images. Since 2015, the IEEE Annual International Low-Power Image Recognition Challenge (LPIRC) has been held to identify the most energy-efficient computer vision solutions. This article summarizes 2018 winners' solutions. (2) Suggest directions for research as well as opportunities for low-power computer vision.

Recommender systems help users deal with information overload by providing tailored item suggestions to them. The recommendation of news is often considered to be challenging, since the relevance of an article for a user can depend on a variety of factors, including the user's short-term reading interests, the reader's context, or the recency or popularity of an article. Previous work has shown that the use of Recurrent Neural Networks is promising for the next-in-session prediction task, but has certain limitations when only recorded item click sequences are used as input. In this work, we present a hybrid, deep learning based approach for session-based news recommendation that is able to leverage a variety of information types. We evaluated our approach on two public datasets, using a temporal evaluation protocol that simulates the dynamics of a news portal in a realistic way. Our results confirm the benefits of considering additional types of information, including article popularity and recency, in the proposed way, resulting in significantly higher recommendation accuracy and catalog coverage than other session-based algorithms. Additional experiments show that the proposed parameterizable loss function used in our method also allows us to balance two usually conflicting quality factors, accuracy and novelty. Keywords: News Recommender Systems, Session-based Recommendation, Artificial Neural Networks, Context-awareness, Hybridization

Can neural networks learn to solve partial differential equations (PDEs)? We investigate this question for two (systems of) PDEs, namely, the Poisson equation and the steady Navier--Stokes equations. The contributions of this paper are five-fold. (1) Numerical experiments show that small neural networks (< 500 learnable parameters) are able to accurately learn complex solutions for systems of partial differential equations. (2) It investigates the influence of random weight initialization on the quality of the neural network approximate solution and demonstrates how one can take advantage of this non-determinism using ensemble learning. (3) It investigates the suitability of the loss function used in this work. (4) It studies the benefits and drawbacks of solving (systems of) PDEs with neural networks compared to classical numerical methods. (5) It proposes an exhaustive list of possible directions of future work.

Deep neural networks have frequently been used to directly learn representations useful for a given task from raw input data. In terms of overall performance metrics, machine learning solutions employing deep representations frequently have been reported to greatly outperform those using hand-crafted feature representations. At the same time, they may pick up on aspects that are predominant in the data, yet not actually meaningful or interpretable. In this paper, we therefore propose a systematic way to diagnose the trustworthiness of deep music representations, considering musical semantics. The underlying assumption is that in case a deep representation is to be trusted, distance consistency between known related points should be maintained both in the input audio space and corresponding latent deep space. We generate known related points through semantically meaningful transformations, both considering imperceptible and graver transformations. Then, we examine within- and between-space distance consistencies, both considering audio space and latent embedded space, the latter either being a result of a conventional feature extractor or a deep encoder. We illustrate how our method, as a complement to task-specific performance, provides interpretable insight into what a network may have captured from training data signals.

The electrocardiogram (ECG) is a widely-used medical test, typically consisting of 12 voltage versus time traces collected from surface recordings over the heart. Here we hypothesize that a deep neural network can predict an important future clinical event (one-year all-cause mortality) from ECG voltage-time traces. We show good performance for predicting one-year mortality with an average AUC of 0.85 from a model cross-validated on 1,775,926 12-lead resting ECGs, that were collected over a 34-year period in a large regional health system. Even within the large subset of ECGs interpreted as 'normal' by a physician (n=297,548), the model performance to predict one-year mortality remained high (AUC=0.84), and Cox Proportional Hazard model revealed a hazard ratio of 6.6 (p<0.005) for the two predicted groups (dead vs alive one year after ECG) over a 30-year follow-up period. A blinded survey of three cardiologists suggested that the patterns captured by the model were generally not visually apparent to cardiologists even after being shown 240 paired examples of labeled true positives (dead) and true negatives (alive). In summary, deep learning can add significant prognostic information to the interpretation of 12-lead resting ECGs, even in cases that are interpreted as 'normal' by physicians.

Globally normalized neural sequence models are considered superior to their locally normalized equivalents because they may ameliorate the effects of label bias. However, when considering high-capacity neural parametrizations that condition on the whole input sequence, both model classes are theoretically equivalent in terms of the distributions they are capable of representing. Thus, the practical advantage of global normalization in the context of modern neural methods remains unclear. In this paper, we attempt to shed light on this problem through an empirical study. We extend an approach for search-aware training via a continuous relaxation of beam search (Goyal et al., 2017b) in order to enable training of globally normalized recurrent sequence models through simple backpropagation. We then use this technique to conduct an empirical study of the interaction between global normalization, high-capacity encoders, and search-aware optimization. We observe that in the context of inexact search, globally normalized neural models are still more effective than their locally normalized counterparts. Further, since our training approach is sensitive to warm-starting with pre-trained models, we also propose a novel initialization strategy based on self-normalization for pre-training globally normalized models. We perform analysis of our approach on two tasks: CCG supertagging and Machine Translation, and demonstrate the importance of global normalization under different conditions while using search-aware training.