Learning Surrogate Losses

TL;DR

This paper introduces a versatile optimization method that learns smooth surrogate losses for non-differentiable metrics, enabling effective training across various complex evaluation criteria in machine learning.

Contribution

It proposes a novel bilevel optimization approach to learn surrogate neural network losses for any non-differentiable evaluation metric, improving training flexibility.

Findings

Effective minimization of diverse real-world loss functions.

Outperforms state-of-the-art baselines on multiple datasets.

Surrogate losses are invariant to mini-batch order.

Abstract

The minimization of loss functions is the heart and soul of Machine Learning. In this paper, we propose an off-the-shelf optimization approach that can minimize virtually any non-differentiable and non-decomposable loss function (e.g. Miss-classification Rate, AUC, F1, Jaccard Index, Mathew Correlation Coefficient, etc.) seamlessly. Our strategy learns smooth relaxation versions of the true losses by approximating them through a surrogate neural network. The proposed loss networks are set-wise models which are invariant to the order of mini-batch instances. Ultimately, the surrogate losses are learned jointly with the prediction model via bilevel optimization. Empirical results on multiple datasets with diverse real-life loss functions compared with state-of-the-art baselines demonstrate the efficiency of learning surrogate losses.