The Gumbel-max trick is a method to draw a sample from a categorical
distribution, given by its unnormalized (log-)probabilities. Over the past
years, the machine learning community has proposed several extensions of this
trick to facilitate, e.g., drawing multiple samples, sampling from structured
domains, or gradient estimation for error backpropagation in neural network
optimization. The goal of this survey article is to present background about
the Gumbel-max trick, and to provide a structured overview of its extensions to
ease algorithm selection. Moreover, it presents a comprehensive outline of
(machine learning) literature in which Gumbel-based algorithms have been
leveraged, reviews commonly-made design choices, and sketches a future
perspective.
Description
A Review of the Gumbel-max Trick and its Extensions for Discrete Stochasticity in Machine Learning
%0 Generic
%1 huijben2021review
%A Huijben, Iris A. M.
%A Kool, Wouter
%A Paulus, Max B.
%A van Sloun, Ruud J. G.
%D 2021
%K PLK differentiable gumbel loss softmax
%T A Review of the Gumbel-max Trick and its Extensions for Discrete
Stochasticity in Machine Learning
%U http://arxiv.org/abs/2110.01515
%X The Gumbel-max trick is a method to draw a sample from a categorical
distribution, given by its unnormalized (log-)probabilities. Over the past
years, the machine learning community has proposed several extensions of this
trick to facilitate, e.g., drawing multiple samples, sampling from structured
domains, or gradient estimation for error backpropagation in neural network
optimization. The goal of this survey article is to present background about
the Gumbel-max trick, and to provide a structured overview of its extensions to
ease algorithm selection. Moreover, it presents a comprehensive outline of
(machine learning) literature in which Gumbel-based algorithms have been
leveraged, reviews commonly-made design choices, and sketches a future
perspective.
@misc{huijben2021review,
abstract = {The Gumbel-max trick is a method to draw a sample from a categorical
distribution, given by its unnormalized (log-)probabilities. Over the past
years, the machine learning community has proposed several extensions of this
trick to facilitate, e.g., drawing multiple samples, sampling from structured
domains, or gradient estimation for error backpropagation in neural network
optimization. The goal of this survey article is to present background about
the Gumbel-max trick, and to provide a structured overview of its extensions to
ease algorithm selection. Moreover, it presents a comprehensive outline of
(machine learning) literature in which Gumbel-based algorithms have been
leveraged, reviews commonly-made design choices, and sketches a future
perspective.},
added-at = {2023-01-10T08:52:10.000+0100},
author = {Huijben, Iris A. M. and Kool, Wouter and Paulus, Max B. and van Sloun, Ruud J. G.},
biburl = {https://www.bibsonomy.org/bibtex/2ba9a2d077502797c3ac69f4d6cdd86ad/parismic},
description = {A Review of the Gumbel-max Trick and its Extensions for Discrete Stochasticity in Machine Learning},
interhash = {a9d48809f8af7fc660f6cb47981bbc41},
intrahash = {ba9a2d077502797c3ac69f4d6cdd86ad},
keywords = {PLK differentiable gumbel loss softmax},
note = {cite arxiv:2110.01515Comment: Accepted as a survey article in IEEE TPAMI},
timestamp = {2023-01-10T13:04:10.000+0100},
title = {A Review of the Gumbel-max Trick and its Extensions for Discrete
Stochasticity in Machine Learning},
url = {http://arxiv.org/abs/2110.01515},
year = 2021
}