A0891
Title: Difficulties and nonstandard minimax rates in nonparametric latent variable models and representation learning
Authors: Bryon Aragam - University of Chicago (United States) [presenting]
Abstract: One of the key paradigm shifts in statistical machine learning over the past decade has been the transition from handcrafted features to automated, data-driven representation learning, typically via deep neural networks. As these methods are being used in high-stakes settings such as medicine, health care, law, and finance, where accountability and transparency are not just desirable but often legally required, it has become necessary to place representation learning on a rigorous scientific footing. The statistical foundations of nonparametric latent variable models are revisited and discussed how even basic statistical properties such as identifiability and consistency are surprisingly subtle. New results are also discussed, characterizing the optimal sample complexity for learning simple nonparametric mixtures, which turn out to have a nonstandard super-polynomial bound. With time permitting, applications will end to deep generative models widely used in practice.
