Estimating Regression Predictive Distributions with Sample Networks

TL;DR

This paper introduces SampleNet, a scalable neural network architecture that models predictive uncertainty using learned samples, avoiding restrictive parametric assumptions and improving fit on diverse distributions in regression tasks.

Contribution

SampleNet provides a novel, flexible approach to uncertainty estimation by learning empirical distributions with energy-based training and divergence regularization.

Findings

SampleNet accurately models a wide range of distributions.

It outperforms baseline methods on large-scale regression tasks.

The approach is scalable and avoids parametric distribution assumptions.

Abstract

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.