DeepAI
Log In Sign Up

Distributionally Robust Losses for Latent Covariate Mixtures

07/28/2020
by   John Duchi, et al.
0

While modern large-scale datasets often consist of heterogeneous subpopulations—for example, multiple demographic groups or multiple text corpora—the standard practice of minimizing average loss fails to guarantee uniformly low losses across all subpopulations. We propose a convex procedure that controls the worst-case performance over all subpopulations of a given size. Our procedure comes with finite-sample (nonparametric) convergence guarantees on the worst-off subpopulation. Empirically, we observe on lexical similarity, wine quality, and recidivism prediction tasks that our worst-case procedure learns models that do well against unseen subpopulations.

READ FULL TEXT
05/26/2021

A data-driven approach to beating SAA out-of-sample

While solutions of Distributionally Robust Optimization (DRO) problems c...
07/26/2020

Beyond the Worst-Case Analysis of Algorithms (Introduction)

One of the primary goals of the mathematical analysis of algorithms is t...
10/12/2021

Balancing Average and Worst-case Accuracy in Multitask Learning

When training and evaluating machine learning models on a large number o...
01/03/2013

Follow the Leader If You Can, Hedge If You Must

Follow-the-Leader (FTL) is an intuitive sequential prediction strategy t...
11/02/2018

Worst-Case Efficient Sorting with QuickMergesort

The two most prominent solutions for the sorting problem are Quicksort a...
07/09/2021

Training Over-parameterized Models with Non-decomposable Objectives

Many modern machine learning applications come with complex and nuanced ...
03/13/2022

Worst Case Matters for Few-Shot Recognition

Few-shot recognition learns a recognition model with very few (e.g., 1 o...