Saved in:
Bibliographic Details
Main Authors: Hopkins, Max, Impagliazzo, Russell, Ye, Christopher
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.20200
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911574135930880
author Hopkins, Max
Impagliazzo, Russell
Ye, Christopher
author_facet Hopkins, Max
Impagliazzo, Russell
Ye, Christopher
contents Replicability, introduced by (Impagliazzo et al. STOC '22), is the notion that algorithms should remain stable under a resampling of their inputs (given access to shared randomness). While a strong and interesting notion of stability, the cost of replicability can be prohibitive: there is no replicable algorithm, for instance, for tasks as simple as threshold learning (Bun et al. STOC '23). Given such strong impossibility results we ask: under what approximate notions of replicability is learning possible? In this work, we propose three natural relaxations of replicability in the context of PAC learning: (1) Pointwise: the learner must be consistent on any fixed input, but not across all inputs simultaneously, (2) Approximate: the learner must output hypotheses that classify most of the distribution consistently, (3) Semi: the algorithm is fully replicable, but may additionally use shared unlabeled samples. In all three cases, for constant replicability we obtain close to sample-optimal agnostic PAC learners: 1) and 2) are achievable using $O(d/α^2 + 1/α^{4})$ samples, while 3) requires $Θ(d^2/α^2)$ labeled samples.
format Preprint
id arxiv_https___arxiv_org_abs_2510_20200
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Approximate Replicability in Learning
Hopkins, Max
Impagliazzo, Russell
Ye, Christopher
Machine Learning
Replicability, introduced by (Impagliazzo et al. STOC '22), is the notion that algorithms should remain stable under a resampling of their inputs (given access to shared randomness). While a strong and interesting notion of stability, the cost of replicability can be prohibitive: there is no replicable algorithm, for instance, for tasks as simple as threshold learning (Bun et al. STOC '23). Given such strong impossibility results we ask: under what approximate notions of replicability is learning possible? In this work, we propose three natural relaxations of replicability in the context of PAC learning: (1) Pointwise: the learner must be consistent on any fixed input, but not across all inputs simultaneously, (2) Approximate: the learner must output hypotheses that classify most of the distribution consistently, (3) Semi: the algorithm is fully replicable, but may additionally use shared unlabeled samples. In all three cases, for constant replicability we obtain close to sample-optimal agnostic PAC learners: 1) and 2) are achievable using $O(d/α^2 + 1/α^{4})$ samples, while 3) requires $Θ(d^2/α^2)$ labeled samples.
title Approximate Replicability in Learning
topic Machine Learning
url https://arxiv.org/abs/2510.20200