Saved in:
Bibliographic Details
Main Authors: Indelman, Hedda Cohen, Hazan, Tamir
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.02180
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917684459864064
author Indelman, Hedda Cohen
Hazan, Tamir
author_facet Indelman, Hedda Cohen
Hazan, Tamir
contents The Gumbel-Softmax probability distribution allows learning discrete tokens in generative learning, while the Gumbel-Argmax probability distribution is useful in learning discrete structures in discriminative learning. Despite the efforts invested in optimizing these probability models, their statistical properties are under-explored. In this work, we investigate their representation properties and determine for which families of parameters these probability distributions are complete, i.e., can represent any probability distribution, and minimal, i.e., can represent a probability distribution uniquely. We rely on convexity and differentiability to determine these statistical conditions and extend this framework to general probability models, such as Gaussian-Softmax and Gaussian-Argmax. We experimentally validate the qualities of these extensions, which enjoy a faster convergence rate. We conclude the analysis by identifying two sets of parameters that satisfy these assumptions and thus admit a complete and minimal representation. Our contribution is theoretical with supporting practical evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2406_02180
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On The Statistical Representation Properties Of The Perturb-Softmax And The Perturb-Argmax Probability Distributions
Indelman, Hedda Cohen
Hazan, Tamir
Machine Learning
Artificial Intelligence
The Gumbel-Softmax probability distribution allows learning discrete tokens in generative learning, while the Gumbel-Argmax probability distribution is useful in learning discrete structures in discriminative learning. Despite the efforts invested in optimizing these probability models, their statistical properties are under-explored. In this work, we investigate their representation properties and determine for which families of parameters these probability distributions are complete, i.e., can represent any probability distribution, and minimal, i.e., can represent a probability distribution uniquely. We rely on convexity and differentiability to determine these statistical conditions and extend this framework to general probability models, such as Gaussian-Softmax and Gaussian-Argmax. We experimentally validate the qualities of these extensions, which enjoy a faster convergence rate. We conclude the analysis by identifying two sets of parameters that satisfy these assumptions and thus admit a complete and minimal representation. Our contribution is theoretical with supporting practical evaluation.
title On The Statistical Representation Properties Of The Perturb-Softmax And The Perturb-Argmax Probability Distributions
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2406.02180