Learning with many experts: model selection and sparsity

TL;DR

This paper introduces a surrogate loss for model selection in noisy expert-labeled data, providing theoretical guarantees and enabling sparsity tuning to improve model interpretability and prevent overfitting.

Contribution

It proposes a new surrogate loss for model selection with noisy labels and demonstrates how to tune sparsity in models under label noise conditions.

Findings

The surrogate loss is consistent for model selection.

Sparsity tuning improves model interpretability.

Method performs well on simulated and real data.

Abstract

Experts classifying data are often imprecise. Recently, several models have been proposed to train classifiers using the noisy labels generated by these experts. How to choose between these models? In such situations, the true labels are unavailable. Thus, one cannot perform model selection using the standard versions of methods such as empirical risk minimization and cross validation. In order to allow model selection, we present a surrogate loss and provide theoretical guarantees that assure its consistency. Next, we discuss how this loss can be used to tune a penalization which introduces sparsity in the parameters of a traditional class of models. Sparsity provides more parsimonious models and can avoid overfitting. Nevertheless, it has seldom been discussed in the context of noisy labels due to the difficulty in model selection and, therefore, in choosing tuning parameters. We apply these techniques to several sets of simulated and real data.