Abstract:  As datasets continue to grow in size, in many settings the focus of data collection has shifted away from testing pre-specified hypotheses, and towards hypothesis generation. Researchers are often interested in performing an exploratory data analysis in order to generate hypotheses, and then testing those hypotheses on the same data; I will refer to this as 'double dipping'. Unfortunately, double dipping can lead to highly-inflated Type 1 errors. In this talk, I will consider the special case of hierarchical…

Abstract: What combinatorial properties are likely to be satisfied by a random subspace over a finite field? For example, is it likely that not too many points lie in any Hamming ball? What about any cube?  We show that there is a sharp threshold on the dimension of the subspace at which the answers to these questions change from "extremely likely" to "extremely unlikely," and moreover we give a simple characterization of this threshold for different properties. Our motivation comes…

Abstract: I will introduce Ensemble Rejection Sampling, a scheme for perfect simulation of a class of Feynmac-Kac models. In particular, this scheme allows us to sample exactly from the posterior distribution of the latent states of a class of non-linear non-Gaussian state-space models and from the distribution of a class of conditioned random walks. Ensemble Rejection Sampling relies on a high-dimensional proposal distribution built using ensembles of state samples and dynamic programming. Although this algorithm can be interpreted as a…

Abstract: The training of neural networks optimizes complex non-convex objective functions, yet in practice simple algorithms achieve great performances. Recent works suggest that over-parametrization could be a key ingredient in explaining this discrepancy. However, current theories could not fully explain the role of over-parameterization. In particular, they either work in a regime where neurons don't move much, or require large number of neurons. In this paper we develop a local convergence theory for mildly over-parameterized two-layer neural net. We show…

Abstract: The goal of list learning is to understand how to learn basic statistics of a dataset when it has been corrupted by an overwhelming fraction of outliers. More formally, one is given a set of points $S$, of which an $\alpha$-fraction $T$ are promised to be well-behaved. The goal is then to output an $O(1 / \alpha)$ sized list of candidate means, so that one of these candidates is close to the true mean of the points in $T$.…

Abstract: Introduced by Kiefer and Wolfowitz 1956, the nonparametric maximum likelihood estimator (NPMLE) is a widely used methodology for learning mixture models and empirical Bayes estimation. Sidestepping the non-convexity in mixture likelihood, the NPMLE estimates the mixing distribution by maximizing the total likelihood over the space of probability measures, which can be viewed as an extreme form of over parameterization. In this work we discover a surprising property of the NPMLE solution. Consider, for example, a Gaussian mixture model on…