Scalable Infomin Learning

TL;DR

This paper introduces a scalable infomin learning method that efficiently learns representations uninformative about specific targets, improving training speed and effectiveness across fairness, disentanglement, and domain adaptation tasks.

Contribution

It proposes a novel proxy metric for mutual information and an analytical approximation, eliminating the need for neural network estimators in infomin learning.

Findings

Effective removal of unwanted information in limited time

Improved fairness and disentanglement in representations

Versatile application across multiple domains

Abstract

The task of infomin learning aims to learn a representation with high utility while being uninformative about a specified target, with the latter achieved by minimising the mutual information between the representation and the target. It has broad applications, ranging from training fair prediction models against protected attributes, to unsupervised learning with disentangled representations. Recent works on infomin learning mainly use adversarial training, which involves training a neural network to estimate mutual information or its proxy and thus is slow and difficult to optimise. Drawing on recent advances in slicing techniques, we propose a new infomin learning approach, which uses a novel proxy metric to mutual information. We further derive an accurate and analytically computable approximation to this proxy metric, thereby removing the need of constructing neural network-based mutual information estimators. Experiments on algorithmic fairness, disentangled representation learning and domain adaptation verify that our method can effectively remove unwanted information with limited time budget.