On the Complexity of Learning Sparse Functions with Statistical and Gradient Queries

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Joshi, Nirmit, Misiakiewicz, Theodor, Srebro, Nathan
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914861419593728
author Joshi, Nirmit
Misiakiewicz, Theodor
Srebro, Nathan
author_facet Joshi, Nirmit
Misiakiewicz, Theodor
Srebro, Nathan
contents The goal of this paper is to investigate the complexity of gradient algorithms when learning sparse functions (juntas). We introduce a type of Statistical Queries ($\mathsf{SQ}$), which we call Differentiable Learning Queries ($\mathsf{DLQ}$), to model gradient queries on a specified loss with respect to an arbitrary model. We provide a tight characterization of the query complexity of $\mathsf{DLQ}$ for learning the support of a sparse function over generic product distributions. This complexity crucially depends on the loss function. For the squared loss, $\mathsf{DLQ}$ matches the complexity of Correlation Statistical Queries $(\mathsf{CSQ})$--potentially much worse than $\mathsf{SQ}$. But for other simple loss functions, including the $\ell_1$ loss, $\mathsf{DLQ}$ always achieves the same complexity as $\mathsf{SQ}$. We also provide evidence that $\mathsf{DLQ}$ can indeed capture learning with (stochastic) gradient descent by showing it correctly describes the complexity of learning with a two-layer neural network in the mean field regime and linear scaling.
format Preprint
id arxiv_https___arxiv_org_abs_2407_05622
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On the Complexity of Learning Sparse Functions with Statistical and Gradient Queries
Joshi, Nirmit
Misiakiewicz, Theodor
Srebro, Nathan
Machine Learning
Data Structures and Algorithms
The goal of this paper is to investigate the complexity of gradient algorithms when learning sparse functions (juntas). We introduce a type of Statistical Queries ($\mathsf{SQ}$), which we call Differentiable Learning Queries ($\mathsf{DLQ}$), to model gradient queries on a specified loss with respect to an arbitrary model. We provide a tight characterization of the query complexity of $\mathsf{DLQ}$ for learning the support of a sparse function over generic product distributions. This complexity crucially depends on the loss function. For the squared loss, $\mathsf{DLQ}$ matches the complexity of Correlation Statistical Queries $(\mathsf{CSQ})$--potentially much worse than $\mathsf{SQ}$. But for other simple loss functions, including the $\ell_1$ loss, $\mathsf{DLQ}$ always achieves the same complexity as $\mathsf{SQ}$. We also provide evidence that $\mathsf{DLQ}$ can indeed capture learning with (stochastic) gradient descent by showing it correctly describes the complexity of learning with a two-layer neural network in the mean field regime and linear scaling.
title On the Complexity of Learning Sparse Functions with Statistical and Gradient Queries
topic Machine Learning
Data Structures and Algorithms
url https://arxiv.org/abs/2407.05622