Generalizing Stochastic Smoothing for Differentiation and Gradient Estimation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Petersen, Felix, Borgelt, Christian, Mishra, Aashwin, Ermon, Stefano
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914969515196416
author Petersen, Felix
Borgelt, Christian
Mishra, Aashwin
Ermon, Stefano
author_facet Petersen, Felix
Borgelt, Christian
Mishra, Aashwin
Ermon, Stefano
contents We deal with the problem of gradient estimation for stochastic differentiable relaxations of algorithms, operators, simulators, and other non-differentiable functions. Stochastic smoothing conventionally perturbs the input of a non-differentiable function with a differentiable density distribution with full support, smoothing it and enabling gradient estimation. Our theory starts at first principles to derive stochastic smoothing with reduced assumptions, without requiring a differentiable density nor full support, and we present a general framework for relaxation and gradient estimation of non-differentiable black-box functions $f:\mathbb{R}^n\to\mathbb{R}^m$. We develop variance reduction for gradient estimation from 3 orthogonal perspectives. Empirically, we benchmark 6 distributions and up to 24 variance reduction strategies for differentiable sorting and ranking, differentiable shortest-paths on graphs, differentiable rendering for pose estimation, as well as differentiable cryo-ET simulations.
format Preprint
id arxiv_https___arxiv_org_abs_2410_08125
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Generalizing Stochastic Smoothing for Differentiation and Gradient Estimation
Petersen, Felix
Borgelt, Christian
Mishra, Aashwin
Ermon, Stefano
Machine Learning
We deal with the problem of gradient estimation for stochastic differentiable relaxations of algorithms, operators, simulators, and other non-differentiable functions. Stochastic smoothing conventionally perturbs the input of a non-differentiable function with a differentiable density distribution with full support, smoothing it and enabling gradient estimation. Our theory starts at first principles to derive stochastic smoothing with reduced assumptions, without requiring a differentiable density nor full support, and we present a general framework for relaxation and gradient estimation of non-differentiable black-box functions $f:\mathbb{R}^n\to\mathbb{R}^m$. We develop variance reduction for gradient estimation from 3 orthogonal perspectives. Empirically, we benchmark 6 distributions and up to 24 variance reduction strategies for differentiable sorting and ranking, differentiable shortest-paths on graphs, differentiable rendering for pose estimation, as well as differentiable cryo-ET simulations.
title Generalizing Stochastic Smoothing for Differentiation and Gradient Estimation
topic Machine Learning
url https://arxiv.org/abs/2410.08125