Black-box optimizers that explore in parameter space have often been shown to outperform more sophisticated action space exploration methods developed specifically for the reinforcement learning problem. We examine these black-box methods closely to identify situations in which they are worse than action space exploration methods and those in which they are superior. Through simple theoretical analyses, we prove that complexity of exploration in parameter space depends on the dimensionality of parameter space, while complexity of exploration in action space depends on both the dimensionality of action space and horizon length. This is also demonstrated empirically by comparing simple exploration methods on several model problems, including Contextual Bandit, Linear Regression and Reinforcement Learning in continuous control.

With the rapid expansion of mobile phone networks in developing countries, large-scale graph machine learning has gained sudden relevance in the study of global poverty. Recent applications range from humanitarian response and poverty estimation to urban planning and epidemic containment. Yet the vast majority of computational tools and algorithms used in these applications do not account for the multi-view nature of social networks: people are related in myriad ways, but most graph learning models treat relations as binary. In this paper, we develop a graph-based convolutional network for learning on multi-view networks. We show that this method outperforms state-of-the-art semi-supervised learning algorithms on three different prediction tasks using mobile phone datasets from three different developing countries. We also show that, while designed specifically for use in poverty research, the algorithm also outperforms existing benchmarks on a broader set of learning tasks on multi-view networks, including node labelling in citation networks.

Nonnegative matrix factorization (NMF) is a linear dimensionality reduction technique for analyzing nonnegative data. A key aspect of NMF is the choice of the objective function that depends on the noise model (or statistics of the noise) assumed on the data. In many applications, the noise model is unknown and difficult to estimate. In this paper, we define a multi-objective NMF (MO-NMF) problem, where several objectives are combined within the same NMF model. We propose to use Lagrange duality to judiciously optimize for a set of weights to be used within the framework of the weighted-sum approach, that is, we minimize a single objective function which is a weighted sum of the all objective functions. We design a simple algorithm using multiplicative updates to minimize this weighted sum. We show how this can be used to find distributionally robust NMF (DR-NMF) solutions, that is, solutions that minimize the largest error among all objectives. We illustrate the effectiveness of this approach on synthetic, document and audio datasets. The results show that DR-NMF is robust to our incognizance of the noise model of the NMF problem.

Experts are already building a future world brimming with artificial intelligence, but here in the present most of us are still trying to figure out what AI even is. This is a technology that will influence many aspects of our lives, from jobs to entertainment to health care, but that also engages with fundamental questions about what it means to be human. Questions like, "what is the nature of creativity?" and "how do we define consciousness?" Posing the question "how can I understand AI?" is nearly as daunting as asking "what is the meaning of life?" In order to help, The Verge has assembled a reading list: a brief but diverse compendium of books, short stories, and blogs, all chosen by leading figures in the AI world to help you better understand artificial intelligence. It's an eclectic selection that ranges from practical primers to Golden Age sci-fi, and while reading everything listed below won't get you a job at Google (though it certainly couldn't hurt), it will give you much-needed context for this confusing and exciting time. So read, enjoy, and get to know the captivating world of AI a little bit better.

We introduce a Bayesian nonparametric regression model for data with multiway (tensor) structure, motivated by an application to periodontal disease (PD) data. Our outcome is the number of diseased sites measured over four different tooth types for each subject, with subject-specific covariates available as predictors. The outcomes are not well-characterized by simple parametric models, so we use a nonparametric approach with a binomial likelihood wherein the latent probabilities are drawn from a mixture with an arbitrary number of components, analogous to a Dirichlet Process (DP). We use a flexible probit stick-breaking formulation for the component weights that allows for covariate dependence and clustering structure in the outcomes. The parameter space for this model is large and multiway: patients $\times$ tooth types $\times$ covariates $\times$ components. We reduce its effective dimensionality, and account for the multiway structure, via low-rank assumptions. We illustrate how this can improve performance, and simplify interpretation, while still providing sufficient flexibility. We describe a general and efficient Gibbs sampling algorithm for posterior computation. The resulting fit to the PD data outperforms competitors, and is interpretable and well-calibrated. An interactive visual of the predictive model is available at http://ericfrazerlock.com/toothdata/ToothDisplay.html , and the code is available at https://github.com/lockEF/NonparametricMultiway .

Recurrent Neural Networks (RNNs) are very successful at solving challenging problems with sequential data. However, this observed efficiency is not yet entirely explained by theory. It is known that a certain class of multiplicative RNNs enjoys the property of depth efficiency --- a shallow network of exponentially large width is necessary to realize the same score function as computed by such an RNN. Such networks, however, are not very often applied to real life tasks. In this work, we attempt to reduce the gap between theory and practice by extending the theoretical analysis to RNNs which employ various nonlinearities, such as Rectified Linear Unit (ReLU), and show that they also benefit from properties of universality and depth efficiency. Our theoretical results are verified by a series of extensive computational experiments.

This week I was invited to speak at the Oxford Union in England on the topic: AI: An inhumane future? In this panel there was Kenneth Cukier, a senior editor of The Economist, D Catherine Havasi, the Massachusetts Institute of Technology (MIT) scientist and Professor Sir Adrian Smith, the director of The Alan Turing Institute. As I was preparing for this talk, I was reminded of the famous British author Charles Dickens who in his classic novel, A Tale of Two Cities, famously characterised the French Revolution as follows: "It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of Light, it was the season of Darkness, it was the spring of hope, it was the winter of despair." Artificial Intelligence (AI) is catalysing another revolution and not the French one…but the fourth industrial revolution (4IR). While the French revolution ushered new improved ways of human relations, it also ushered the cruelty of political factions.

Climate change is the most horrific threat our species has ever known: No matter how powerful you are or how much money you have, our transforming planet is a reckoning for every one of us. But there are degrees to this misery. If you're perched in a Manhattan penthouse, the effects might not be immediately apparent (because you don't care or aren't paying attention, or both). If you're a subsistence farmer in Kenya, the situation is already much more dire. There's an equalizer, though, that is helping small farmers adapt to a changing planet: smartphones.

On the occasion of Data Protection Day on 28 January, the Committee of the Council of Europe's data protection treaty "Convention 108" has published Guidelines on Artificial Intelligence and Data Protection. The guidelines aim to assist policy makers, artificial intelligence (AI) developers, manufacturers and service providers in ensuring that AI applications do not undermine the right to data protection. The Convention's Committee underlines that the protection of human rights, including the right to protection of personal data, should be an essential pre-requisite when developing or adopting AI applications, in particular when they are used in decision-making processes, and be based on the principles of the updated data protection convention, "Convention 108, opened for signature in October 2018. In addition, any innovation in the field of AI should pay close attention to avoiding and mitigating the potential risks of processing of personal data, and allow meaningful control by data subjects over the data processing and its effects. Minister for Foreign Affairs of Finland and Chair of the Committee of Ministers of the Council of Europe Timo Soini welcomed the adoption of the guidelines and said: "Artificial intelligence brings benefits to our daily lives.

TL;DR: I had a pleasure to be a part of the first AI conference in Somaliland, organised by Shaqodoon, HarHub and Elmi Academy, and featuring speakers from Google, MIT, major Somaliland telecoms, banks, University of Hargeisa and Ministry of Telecommunication & Technology of Somaliland. See slides (lectures workshops) and event program for details. Below are my personal notes and pictures from this trip. Whenever I tell this story people seem to be surprised with choice of spending vacation time in Somaliland and running an AI-related event there. So let me share some first-hand experience with you and explain why trips and events like this are useful, fun and safe.