AutoTransfer: Subject Transfer Learning with Censored Representations on Biosignals Data

TL;DR

AutoTransfer introduces a flexible regularization framework for subject transfer learning on biosignals, leveraging independence penalties to improve model generalization across subjects in EEG, EMG, and ECoG data.

Contribution

The paper proposes a novel regularization framework with multiple independence penalties and an automated strategy for applying them to enhance transfer learning on biosignals.

Findings

Improved subject transfer performance on EEG, EMG, and ECoG datasets.

Effective estimation algorithms for independence penalties.

AutoTransfer simplifies applying regularization to new datasets.

Abstract

We provide a regularization framework for subject transfer learning in which we seek to train an encoder and classifier to minimize classification loss, subject to a penalty measuring independence between the latent representation and the subject label. We introduce three notions of independence and corresponding penalty terms using mutual information or divergence as a proxy for independence. For each penalty term, we provide several concrete estimation algorithms, using analytic methods as well as neural critic functions. We provide a hands-off strategy for applying this diverse family of regularization algorithms to a new dataset, which we call "AutoTransfer". We evaluate the performance of these individual regularization strategies and our AutoTransfer method on EEG, EMG, and ECoG datasets, showing that these approaches can improve subject transfer learning for challenging real-world datasets.