Scalable Infomin Learning
This work addresses the problem of slow and difficult optimization in infomin learning for applications like fair prediction and unsupervised learning, offering a more efficient solution.
The paper tackles the challenge of infomin learning, which aims to learn representations that are useful but uninformative about a specified target, by proposing a new approach that uses a novel proxy metric for mutual information and an analytically computable approximation, eliminating the need for neural network-based estimators. Experiments show the method effectively removes unwanted information with limited time budget, achieving results in fairness, disentanglement, and domain adaptation.
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.