Estimating Regression Predictive Distributions with Sample Networks

Estimating the uncertainty in deep neural network predictions is crucial for many real-world applications. A common approach to model uncertainty is to choose a parametric distribution and fit the data to it using maximum likelihood estimation. The chosen parametric form can be a poor fit to the data-generating distribution, resulting in unreliable uncertainty estimates. In this work, we propose SampleNet, a flexible and scalable architecture for modeling uncertainty that avoids specifying a parametric form on the output distribution. SampleNets do so by defining an empirical distribution using samples that are learned with the Energy Score and regularized with the Sinkhorn Divergence. SampleNets are shown to be able to well-fit a wide range of distributions and to outperform baselines on large-scale real-world regression tasks.

Authors

• Ali Harakeh Mila - Quebec AI Institute Université de Montréal
• Jordan Sir Kwang Hu University of Toronto Institute for Aerospace Studies University of Toronto
• Naiqing Guan University of Toronto
• Steven Waslander University of Toronto Institute for Aerospace Studies University of Toronto
• Liam Paull Mila - Quebec AI Institute Université de Montréal

Keywords

• CV: 3D Computer Vision
• ML: Bayesian Learning
• ML: Probabilistic Methods