A General Method for Robust Learning from Batches

Abstract

In many applications, data is collected in batches, some of which are corrupt or even adversarial. Recent work derived optimal robust algorithms for estimating discrete distributions in this setting. We consider a general framework of robust learning from batches, and determine the limits of both classification and distribution estimation over arbitrary, including continuous, domains. Building on these results, we derive the first robust agnostic computationally-efficient learning algorithms for piecewise-interval classification, and for piecewise-polynomial, monotone, log-concave, and gaussian-mixture distribution estimation.

First Draft

References in corpus (5)

• Communication-Efficient Learning of Deep Networks from Decentralized Data
• Resilience: A Criterion for Learning in the Presence of Arbitrary Outliers
• Learning Discrete Distributions from Untrusted Batches
• Optimal Robust Learning of Discrete Distributions from Batches
• Some techniques in density estimation

Cited by in corpus (2)

• Learning Structured Distributions From Untrusted Batches: Faster and Simpler
• Robust Estimation for Random Graphs