Efficient PAC Learning of Halfspaces with Constant Malicious Noise Rate

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Shen, Jie
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912232680456192
author Shen, Jie
author_facet Shen, Jie
contents Understanding noise tolerance of machine learning algorithms is a central quest in learning theory. In this work, we study the problem of computationally efficient PAC learning of halfspaces in the presence of malicious noise, where an adversary can corrupt both instances and labels of training samples. The best-known noise tolerance either depends on a target error rate under distributional assumptions or on a margin parameter under large-margin conditions. In this work, we show that when both types of conditions are satisfied, it is possible to achieve constant noise tolerance by minimizing a reweighted hinge loss. Our key ingredients include: 1) an efficient algorithm that finds weights to control the gradient deterioration from corrupted samples, and 2) a new analysis on the robustness of the hinge loss equipped with such weights.
format Preprint
id arxiv_https___arxiv_org_abs_2410_01186
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Efficient PAC Learning of Halfspaces with Constant Malicious Noise Rate
Shen, Jie
Machine Learning
Data Structures and Algorithms
Understanding noise tolerance of machine learning algorithms is a central quest in learning theory. In this work, we study the problem of computationally efficient PAC learning of halfspaces in the presence of malicious noise, where an adversary can corrupt both instances and labels of training samples. The best-known noise tolerance either depends on a target error rate under distributional assumptions or on a margin parameter under large-margin conditions. In this work, we show that when both types of conditions are satisfied, it is possible to achieve constant noise tolerance by minimizing a reweighted hinge loss. Our key ingredients include: 1) an efficient algorithm that finds weights to control the gradient deterioration from corrupted samples, and 2) a new analysis on the robustness of the hinge loss equipped with such weights.
title Efficient PAC Learning of Halfspaces with Constant Malicious Noise Rate
topic Machine Learning
Data Structures and Algorithms
url https://arxiv.org/abs/2410.01186