Learning with Algorithmic Supervision via Continuous Relaxations

Felix Petersen, Christian Borgelt, Hilde Kühne, Oliver Deussen

Research output: Chapter in Book/Report/Conference proceeding/Legal commentaryConference contributionpeer-review

Abstract

The integration of algorithmic components into neural architectures has gained increased attention recently, as it allows training neural networks with new forms of supervision such as ordering constraints or silhouettes instead of using ground truth labels. Many approaches in the field focus on the continuous relaxation of a specific task and show promising results in this context. But the focus on single tasks also limits the applicability of the proposed concepts to a narrow range of applications. In this work, we build on those ideas to propose an approach that allows to integrate algorithms into end-to-end trainable neural network architectures based on a general approximation of discrete conditions. To this end, we relax these conditions in control structures such as conditional statements, loops, and indexing, so that resulting algorithms are smoothly differentiable. To obtain meaningful gradients, each relevant variable is perturbed via logistic distributions and the expectation value under this perturbation is approximated. We evaluate the proposed continuous relaxation model on four challenging tasks and show that it can keep up with relaxations specifically designed for each individual task.
Original languageEnglish
Title of host publicationProceedings of the 35th Conference of the Neural Information Processing Society
Publication statusPublished - 7 Dec 2021
Event35th Conference of the Neural Information Processing Society - Online
Duration: 7 Dec 202114 Dec 2021
https://nips.cc/Conferences/2021

Publication series

NameProceedings of the Conference of the Neural Information Processing Society

Conference

Conference35th Conference of the Neural Information Processing Society
Abbreviated titleNeurIPS 2021
Period7/12/2114/12/21
Internet address

Fields of Science and Technology Classification 2012

  • 102 Computer Sciences

Cite this