Learning multivariate Gaussians with imperfect advice

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bhattacharyya, Arnab, Choo, Davin, John, Philips George, Gouleakis, Themis
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910806597173248
author Bhattacharyya, Arnab
Choo, Davin
John, Philips George
Gouleakis, Themis
author_facet Bhattacharyya, Arnab
Choo, Davin
John, Philips George
Gouleakis, Themis
contents We revisit the problem of distribution learning within the framework of learning-augmented algorithms. In this setting, we explore the scenario where a probability distribution is provided as potentially inaccurate advice on the true, unknown distribution. Our objective is to develop learning algorithms whose sample complexity decreases as the quality of the advice improves, thereby surpassing standard learning lower bounds when the advice is sufficiently accurate. Specifically, we demonstrate that this outcome is achievable for the problem of learning a multivariate Gaussian distribution $N(\boldsymbolμ, \boldsymbolΣ)$ in the PAC learning setting. Classically, in the advice-free setting, $\tildeΘ(d^2/\varepsilon^2)$ samples are sufficient and worst case necessary to learn $d$-dimensional Gaussians up to TV distance $\varepsilon$ with constant probability. When we are additionally given a parameter $\tilde{\boldsymbolΣ}$ as advice, we show that $\tilde{O}(d^{2-β}/\varepsilon^2)$ samples suffices whenever $\| \tilde{\boldsymbolΣ}^{-1/2} \boldsymbolΣ \tilde{\boldsymbolΣ}^{-1/2} - \boldsymbol{I_d} \|_1 \leq \varepsilon d^{1-β}$ (where $\|\cdot\|_1$ denotes the entrywise $\ell_1$ norm) for any $β> 0$, yielding a polynomial improvement over the advice-free setting.
format Preprint
id arxiv_https___arxiv_org_abs_2411_12700
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning multivariate Gaussians with imperfect advice
Bhattacharyya, Arnab
Choo, Davin
John, Philips George
Gouleakis, Themis
Machine Learning
Data Structures and Algorithms
Information Theory
We revisit the problem of distribution learning within the framework of learning-augmented algorithms. In this setting, we explore the scenario where a probability distribution is provided as potentially inaccurate advice on the true, unknown distribution. Our objective is to develop learning algorithms whose sample complexity decreases as the quality of the advice improves, thereby surpassing standard learning lower bounds when the advice is sufficiently accurate. Specifically, we demonstrate that this outcome is achievable for the problem of learning a multivariate Gaussian distribution $N(\boldsymbolμ, \boldsymbolΣ)$ in the PAC learning setting. Classically, in the advice-free setting, $\tildeΘ(d^2/\varepsilon^2)$ samples are sufficient and worst case necessary to learn $d$-dimensional Gaussians up to TV distance $\varepsilon$ with constant probability. When we are additionally given a parameter $\tilde{\boldsymbolΣ}$ as advice, we show that $\tilde{O}(d^{2-β}/\varepsilon^2)$ samples suffices whenever $\| \tilde{\boldsymbolΣ}^{-1/2} \boldsymbolΣ \tilde{\boldsymbolΣ}^{-1/2} - \boldsymbol{I_d} \|_1 \leq \varepsilon d^{1-β}$ (where $\|\cdot\|_1$ denotes the entrywise $\ell_1$ norm) for any $β> 0$, yielding a polynomial improvement over the advice-free setting.
title Learning multivariate Gaussians with imperfect advice
topic Machine Learning
Data Structures and Algorithms
Information Theory
url https://arxiv.org/abs/2411.12700