ANC Seminar - Eric Nalisnick

Speaker: Eric Nalisnick

Title:  On Prior Specification for Bayesian Neural Networks

Abstract:  Specifying a Bayesian prior is notoriously difficult for complex models such as neural networks. Reasoning about parameters is made challenging by the high-dimensionality and over-parameterization of the space. Priors that seem benign and uninformative can have unintuitive and detrimental effects on a model’s predictions. To help cope with these problems, I will describe our work on predictive complexity priors: a prior that is defined by comparing the model’s predictions to those of a reference model.  I will show applications to high-dimensional regression, reasoning over neural network depth, and sharing of statistical strength for few-shot learning.