All Random Features Representations are Equivalent

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sernau, Luke, Bonacina, Silvano, Saurous, Rif A.
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912083908493312
author Sernau, Luke
Bonacina, Silvano
Saurous, Rif A.
author_facet Sernau, Luke
Bonacina, Silvano
Saurous, Rif A.
contents Random features are a powerful technique for rewriting positive-definite kernels as linear products. They bring linear tools to bear in important nonlinear domains like KNNs and attention. Unfortunately, practical implementations require approximating an expectation, usually via sampling. This has led to the development of increasingly elaborate representations with ever lower sample error. We resolve this arms race by deriving an optimal sampling policy. Under this policy all random features representations have the same approximation error, which we show is the lowest possible. This means that we are free to choose whatever representation we please, provided we sample optimally.
format Preprint
id arxiv_https___arxiv_org_abs_2406_18802
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle All Random Features Representations are Equivalent
Sernau, Luke
Bonacina, Silvano
Saurous, Rif A.
Machine Learning
Artificial Intelligence
Random features are a powerful technique for rewriting positive-definite kernels as linear products. They bring linear tools to bear in important nonlinear domains like KNNs and attention. Unfortunately, practical implementations require approximating an expectation, usually via sampling. This has led to the development of increasingly elaborate representations with ever lower sample error. We resolve this arms race by deriving an optimal sampling policy. Under this policy all random features representations have the same approximation error, which we show is the lowest possible. This means that we are free to choose whatever representation we please, provided we sample optimally.
title All Random Features Representations are Equivalent
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2406.18802