Training a Subsampling Mechanism in Expectation

Colin Raffel, Dieterich Lawson

Abstract: We describe a mechanism for subsampling sequences and show how to compute its expected output so that it can be trained with standard backpropagation. We test this approach on a simple toy problem and discuss its shortcomings.

TL;DR: We describe a mechanism for subsampling sequences which can be trained in expectation with standard backpropagation.

Conflicts: google.com

Keywords: Theory, Deep learning, Structured prediction