Agnostic Learnability of Halfspaces via Logistic Loss

TL;DR

This paper studies the limits of logistic regression in agnostic learning of halfspaces, establishing tight bounds on its approximation guarantees and proposing a simpler two-step convex optimization method.

Contribution

It constructs a distribution showing the tightness of existing bounds and introduces a new, simpler two-step convex optimization approach to achieve near-optimal risk.

Findings

Constructed a distribution matching the lower bound of $ ilde{ ext{OPT}}$ for logistic regression.

Proposed a two-step convex optimization algorithm that attains $ ilde{O}( ext{OPT})$ misclassification risk.

Demonstrated conditions under which logistic regression reaches near-optimal performance.

Abstract

We investigate approximation guarantees provided by logistic regression for the fundamental problem of agnostic learning of homogeneous halfspaces. Previously, for a certain broad class of "well-behaved" distributions on the examples, Diakonikolas et al. (2020) proved an $\tilde{\Omega}(\textrm{OPT})$ lower bound, while Frei et al. (2021) proved an $\tilde{O}(\sqrt{\textrm{OPT}})$ upper bound, where $\textrm{OPT}$ denotes the best zero-one/misclassification risk of a homogeneous halfspace. In this paper, we close this gap by constructing a well-behaved distribution such that the global minimizer of the logistic risk over this distribution only achieves $\Omega(\sqrt{\textrm{OPT}})$ misclassification risk, matching the upper bound in (Frei et al., 2021). On the other hand, we also show that if we impose a radial-Lipschitzness condition in addition to well-behaved-ness on the distribution, logistic regression on a ball of bounded radius reaches $\tilde{O}(\textrm{OPT})$ misclassification risk. Our techniques also show for any well-behaved distribution, regardless of radial Lipschitzness, we can overcome the $\Omega(\sqrt{\textrm{OPT}})$ lower bound for logistic loss simply at the cost of one additional convex optimization step involving the hinge loss and attain $\tilde{O}(\textrm{OPT})$ misclassification risk. This two-step convex optimization algorithm is simpler than previous methods obtaining this guarantee, all of which require solving $O(\log(1/\textrm{OPT}))$ minimization problems.