Suppose an agent is in a (possibly unknown) Markov decision process (MDP) in the absence of a reward signal, what might we hope that an agent can efficiently learn to do? One natural, intrinsically defined, objective problem is for the agent to learn a policy which induces a distribution over state space that is as uniform as possible, which can be measured in an entropic sense. Despite the corresponding mathematical program being non-convex, our main result provides a provably efficient method (both in terms of sample size and computational complexity) to construct such a maximum-entropy exploratory policy. Key to our algorithmic methodology is utilizing the conditional gradient method (a.k.a. the Frank-Wolfe algorithm) which utilizes an approximate MDP solver.

In the brain, learning signals change over time and synaptic location, and are applied based on the learning history at the synapse, in the complex process of neuromodulation. Learning in artificial neural networks, on the other hand, is shaped by hyper-parameters set before learning starts, which remain static throughout learning, and which are uniform for the entire network. In this work, we propose a method of deep artificial neuromodulation which applies the concepts of biological neuromodulation to stochastic gradient descent. Evolved neuromodulatory dynamics modify learning parameters at each layer in a deep neural network over the course of the network's training. We show that the same neuromodulatory dynamics can be applied to different models and can scale to new problems not encountered during evolution. Finally, we examine the evolved neuromodulation, showing that evolution found dynamic, location-specific learning strategies.

While machine learning is traditionally a resource intensive task, embedded systems, autonomous navigation and the vision of the Internet-of-Things fuel the interest in resource efficient approaches. These approaches require a carefully chosen trade-off between performance and resource consumption in terms of computation and energy. On top of this, it is crucial to treat uncertainty in a consistent manner in all but the simplest applications of machine learning systems. In particular, a desideratum for any real-world system is to be robust in the presence of outliers and corrupted data, as well as being `aware' of its limits, i.e.\ the system should maintain and provide an uncertainty estimate over its own predictions. These complex demands are among the major challenges in current machine learning research and key to ensure a smooth transition of machine learning technology into every day's applications. In this article, we provide an overview of the current state of the art of machine learning techniques facilitating these real-world requirements. First we provide a comprehensive review of resource-efficiency in deep neural networks with focus on techniques for model size reduction, compression and reduced precision. These techniques can be applied during training or as post-processing and are widely used to reduce both computational complexity and memory footprint. As most (practical) neural networks are limited in their ways to treat uncertainty, we contrast them with probabilistic graphical models, which readily serve these desiderata by means of probabilistic inference. In that way, we provide an extensive overview of the current state-of-the-art of robust and efficient machine learning for real-world systems.

We present an off-policy actor-critic algorithm for Reinforcement Learning (RL) that combines ideas from gradient-free optimization via stochastic search with learned action-value function. The result is a simple procedure consisting of three steps: i) policy evaluation by estimating a parametric action-value function; ii) policy improvement via the estimation of a local non-parametric policy; and iii) generalization by fitting a parametric policy. Each step can be implemented in different ways, giving rise to several algorithm variants. Our algorithm draws on connections to existing literature on black-box optimization and 'RL as an inference' and it can be seen either as an extension of the Maximum a Posteriori Policy Optimisation algorithm (MPO) [Abdolmaleki et al., 2018a], or as an extension of Trust Region Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES) [Abdolmaleki et al., 2017b; Hansen et al., 1997] to a policy iteration scheme. Our comparison on 31 continuous control tasks from parkour suite [Heess et al., 2017], DeepMind control suite [Tassa et al., 2018] and OpenAI Gym [Brockman et al., 2016] with diverse properties, limited amount of compute and a single set of hyperparameters, demonstrate the effectiveness of our method and the state of art results. Videos, summarizing results, can be found at goo.gl/HtvJKR .

This paper considers extractive summarisation in a comparative setting: given two or more document groups (e.g., separated by publication time), the goal is to select a small number of documents that are representative of each group, and also maximally distinguishable from other groups. We formulate a set of new objective functions for this problem that connect recent literature on document summarisation, interpretable machine learning, and data subset selection. In particular, by casting the problem as a binary classification amongst different groups, we derive objectives based on the notion of maximum mean discrepancy, as well as a simple yet effective gradient-based optimisation strategy. Our new formulation allows scalable evaluations of comparative summarisation as a classification task, both automatically and via crowd-sourcing. To this end, we evaluate comparative summarisation methods on a newly curated collection of controversial news topics over 13 months. We observe that gradient-based optimisation outperforms discrete and baseline approaches in 15 out of 24 different automatic evaluation settings. In crowd-sourced evaluations, summaries from gradient optimisation elicit 7% more accurate classification from human workers than discrete optimisation. Our result contrasts with recent literature on submodular data subset selection that favours discrete optimisation. We posit that our formulation of comparative summarisation will prove useful in a diverse range of use cases such as comparing content sources, authors, related topics, or distinct view points.

The Earth Mover's Distance (EMD) is a state-of-the art metric for comparing probability distributions. The high distinguishability offered by the EMD comes at a high cost in computational complexity. Therefore, linear-complexity approximation algorithms have been proposed to improve its scalability. However, these algorithms are either limited to vector spaces with only a few dimensions or require the probability distributions to populate the vector space sparsely. We propose novel approximation algorithms that overcome both of these limitations, yet still achieve linear time complexity. All our algorithms are data parallel, and therefore, we can take advantage of massively parallel computing engines, such as Graphics Processing Units (GPUs). The experiments on MNIST images show that the new algorithms can perform a billion distance computations in less than a minute using a single GPU. On the popular text-based 20 Newsgroups dataset, the new algorithms are four orders of magnitude faster than the state-of-the-art FastEMD library and match its search accuracy.

In the era of big data, an important weapon in a machine learning researcher's arsenal is a scalable Support Vector Machine (SVM) algorithm. SVMs are extensively used for solving classification problems. Traditional algorithms for learning SVMs often scale super linearly with training set size which becomes infeasible very quickly for large data sets. In recent years, scalable algorithms have been designed which study the primal or dual formulations of the problem. This often suggests a way to decompose the problem and facilitate development of distributed algorithms. In this paper, we present a distributed algorithm for learning linear Support Vector Machines in the primal form for binary classification called Gossip-bAseD sub-GradiEnT (GADGET) SVM. The algorithm is designed such that it can be executed locally on nodes of a distributed system. Each node processes its local homogeneously partitioned data and learns a primal SVM model. It then gossips with random neighbors about the classifier learnt and uses this information to update the model. Extensive theoretical and empirical results suggest that this anytime algorithm has performance comparable to its centralized and online counterparts.

Modeling a sequence of interactions between users and items (e.g., products, posts, or courses) is crucial in domains such as e-commerce, social networking, and education to predict future interactions. Representation learning presents an attractive solution to model the dynamic evolution of user and item properties, where each user/item can be embedded in a euclidean space and its evolution can be modeled by dynamic changes in embedding. However, existing embedding methods either generate static embeddings, treat users and items independently, or are not scalable. Here we present JODIE, a coupled recurrent model to jointly learn the dynamic embeddings of users and items from a sequence of user-item interactions. JODIE has three components. First, the update component updates the user and item embedding from each interaction using their previous embeddings with the two mutually-recursive Recurrent Neural Networks. Second, a novel projection component is trained to forecast the embedding of users at any future time. Finally, the prediction component directly predicts the embedding of the item in a future interaction. For models that learn from a sequence of interactions, traditional training data batching cannot be done due to complex user-user dependencies. Therefore, we present a novel batching algorithm called t-Batch that generates time-consistent batches of training data that can run in parallel, giving massive speed-up. We conduct six experiments on two prediction tasks---future interaction prediction and state change prediction---using four real-world datasets. We show that JODIE outperforms six state-of-the-art algorithms in these tasks by up to 22.4%. Moreover, we show that JODIE is highly scalable and up to 9.2x faster than comparable models. As an additional experiment, we illustrate that JODIE can predict student drop-out from courses five interactions in advance.

We formulate coherence modeling as a regression task and propose two novel methods to combine techniques from our setup with pairwise approaches. The first of our methods is a model that we call "first-next," which operates similarly to selection sorting but conditions decision-making on information about already-sorted sentences. The second consists of a technique for adding context to regression-based models by concatenating sentence-level representations with an encoding of its corresponding out-of-order paragraph. This latter model achieves Kendall-tau distance and positional accuracy scores that match or exceed the current state-of-the-art on these metrics. Our results suggest that many of the gains that come from more complex, machine-translation inspired approaches can be achieved with simpler, more efficient models.

Large scale astronomical surveys continue to increase their depth and scale, providing new opportunities to observe large numbers of celestial objects with ever increasing precision. At the same time, the sheer scale of ongoing and future surveys pose formidable challenges to classify astronomical objects. Pioneering efforts on this front include the citizen science approach adopted by the Sloan Digital Sky Survey (SDSS). These SDSS datasets have been used recently to train neural network models to classify galaxies in the Dark Energy Survey (DES) that overlap the footprint of both surveys. While this represents a significant step to classify unlabeled images of astrophysical objects in DES, the key issue at heart still remains, i.e., the classification of unlabelled DES galaxies that have not been observed in previous surveys. To start addressing this timely and pressing matter, we demonstrate that knowledge from deep learning algorithms trained with real-object images can be transferred to classify elliptical and spiral galaxies that overlap both SDSS and DES surveys, achieving state-of-the-art accuracy 99.6%. More importantly, to initiate the characterization of unlabelled DES galaxies that have not been observed in previous surveys, we demonstrate that our neural network model can also be used for unsupervised clustering, grouping together unlabeled DES galaxies into spiral and elliptical types. We showcase the application of this novel approach by classifying over ten thousand unlabelled DES galaxies into spiral and elliptical classes. We conclude by showing that unsupervised clustering can be combined with recursive training to start creating large-scale DES galaxy catalogs in preparation for the Large Synoptic Survey Telescope era.