This paper develops new methods to recover the missing entries of a high-rank or even full-rank matrix when the intrinsic dimension of the data is low compared to the ambient dimension. Specifically, we assume that the columns of a matrix are generated by polynomials acting on a low-dimensional intrinsic variable, and wish to recover the missing entries under this assumption. We show that we can identify the complete matrix of minimum intrinsic dimension by minimizing the rank of the matrix in a high dimensional feature space. We develop a new formulation of the resulting problem using the kernel trick together with a new relaxation of the rank objective, and propose an efficient optimization method. We also show how to use our methods to complete data drawn from multiple nonlinear manifolds. Comparative studies on synthetic data, subspace clustering with missing data, motion capture data recovery, and transductive learning verify the superiority of our methods over the state-of-the-art.

We study the generalization properties of minimum-norm solutions for three over-parametrized machine learning models including the random feature model, the two-layer neural network model and the residual network model. We proved that for all three models, the generalization error for the minimum-norm solution is comparable to the Monte Carlo rate, up to some logarithmic terms, as long as the models are sufficiently over-parametrized.

While sample sizes in randomized clinical trials are large enough to estimate the average treatment effect well, they are often insufficient for estimation of treatment-covariate interactions critical to studying data-driven precision medicine. Observational data from real world practice may play an important role in alleviating this problem. One common approach in trials is to predict the outcome of interest with separate regression models in each treatment arm, and recommend interventions based on the contrast of the predicted outcomes. Unfortunately, this simple approach may induce spurious treatment-covariate interaction in observational studies when the regression model is misspecified. Motivated by the need of modeling the number of relapses in multiple sclerosis patients, where the ratio of relapse rates is a natural choice of the treatment effect, we propose to estimate the conditional average treatment effect (CATE) as the relative ratio of the potential outcomes, and derive a doubly robust estimator of this CATE in a semiparametric model of treatment-covariate interactions. We also provide a validation procedure to check the quality of the estimator on an independent sample. We conduct simulations to demonstrate the finite sample performance of the proposed methods, and illustrate the advantage of this approach on real data examining the treatment effect of dimethyl fumarate compared to teriflunomide in multiple sclerosis patients.

Pseudo-Random Numbers Generators (PRNGs) are algorithms produced to generate long sequences of statistically uncorrelated numbers, i.e. Pseudo-Random Numbers (PRNs). These numbers are widely employed in mid-level cryptography and in software applications. Test suites are used to evaluate PRNGs quality by checking statistical properties of the generated sequences. Machine learning techniques are often used to break these generators, for instance approximating a certain generator or a certain sequence using a neural network. But what about using machine learning to generate PRNs generators? This paper proposes a Reinforcement Learning (RL) approach to the task of generating PRNGs from scratch by learning a policy to solve an N-dimensional navigation problem. In this context, N is the length of the period of the generated sequence, and the policy is iteratively improved using the average value of an appropriate test suite run over that period. Aim of this work is to demonstrate the feasibility of the proposed approach, to compare it with classical methods, and to lay the foundation of a research path which combines RL and PRNGs.

Detecting biases in artificial intelligence has become difficult because of the impenetrable nature of deep learning. The central difficulty is in relating unobservable phenomena deep inside models with observable, outside quantities that we can measure from inputs and outputs. For example, can we detect gendered perceptions of occupations (e.g., female librarian, male electrician) using questions to and answers from a word embedding-based system? Current techniques for detecting biases are often customized for a task, dataset, or method, affecting their generalization. In this work, we draw from Psychophysics in Experimental Psychology---meant to relate quantities from the real world (i.e., "Physics") into subjective measures in the mind (i.e., "Psyche")---to propose an intellectually coherent and generalizable framework to detect biases in AI. Specifically, we adapt the two-alternative forced choice task (2AFC) to estimate potential biases and the strength of those biases in black-box models. We successfully reproduce previously-known biased perceptions in word embeddings and sentiment analysis predictions. We discuss how concepts in experimental psychology can be naturally applied to understanding artificial mental phenomena, and how psychophysics can form a useful methodological foundation to study fairness in AI.

We consider the problem of learning generalized first-order representations of concepts from a single example. To address this challenging problem, we augment an inductive logic programming learner with two novel algorithmic contributions. First, we define a distance measure between candidate concept representations that improves the efficiency of search for target concept and generalization. Second, we leverage richer human inputs in the form of advice to improve the sample-efficiency of learning. We prove that the proposed distance measure is semantically valid and use that to derive a P AC bound. Our experimental analysis on diverse concept learning tasks demonstrates both the effectiveness and efficiency of the proposed approach over a first-order concept learner using only examples.

One of the things that really annoys AI researchers is how supposedly "intelligent" machines are judged by much higher standards than are humans. Take self-driving cars, they say. So far they've driven millions of miles with very few accidents, a tiny number of them fatal. Yet whenever an autonomous vehicle kills someone there's a huge hoo-ha, while every year in the US nearly 40,000 people die in crashes involving conventional vehicles. Likewise, the AI evangelists complain, everybody and his dog (this columnist included) is up in arms about algorithmic bias: the way in which automated decision-making systems embody the racial, gender and other prejudices implicit in the data sets on which they were trained.

This is an easy prediction to make, because even Tesla isn't claiming that its eye-catching angular steel beast will be available for sale in 2020. The company's own pitch is that production won't even begin until 2021, with owners receiving their first shipments in 2022. But the gap is relevant to Tesla's future: where the company was once genuinely ahead of the curve, in making beautiful electric cars that people wanted to buy, it has increasingly relied on beating its competitors to announcements, rather than actually shipping. The list of Elon Musk's as-yet-unfulfilled promises grows every year – but the electric fleets of BMW, Ford, General Motors and others grow faster. One of the most impressive, and futuristic, products to have come from Google in recent years, Duplex is an AI assistant that can make calls to local businesses on your behalf to do things like book appointments and find out opening times.

The world's largest AI research conference is underway in Vancouver, Canada. Researchers are presenting more than 1,400 papers at the Neural Information Processing Systems (NeurIPS) conference, ranging from work that organizers believe has had the greatest impact over the past decade to Yoshua Bengio's continued march toward consciousness for deep learning. But even as the conference showed theoretical research and neuroscience-related papers on the rise alongside categories like algorithms and deep learning, the mushrooming of the event itself -- and the associated growing pains -- was a constant theme, and it speaks to the growth of the AI field in general. Organizers said that at the start of the conference Sunday, they expected about 400 people to show up for registration. All told, NeurIPS 2019 welcomed 13,000 attendees, up 40% from the prior year.