Rethinking Importance Weighting for Transfer Learning

TL;DR

This paper reviews recent advances in transfer learning that address distribution shifts, focusing on importance-weighting, joint and dynamic importance estimation, and causal structure transfer, highlighting new methods for complex, high-dimensional tasks.

Contribution

It introduces a comprehensive review of recent transfer learning methods, including novel approaches like causal mechanism transfer and dynamic importance estimation.

Findings

Importance-weighting underpins many transfer learning methods.

Recent advances include joint and dynamic importance estimation techniques.

Causal structure transfer offers new avenues for transfer learning.

Abstract

A key assumption in supervised learning is that training and test data follow the same probability distribution. However, this fundamental assumption is not always satisfied in practice, e.g., due to changing environments, sample selection bias, privacy concerns, or high labeling costs. Transfer learning (TL) relaxes this assumption and allows us to learn under distribution shift. Classical TL methods typically rely on importance-weighting -- a predictor is trained based on the training losses weighted according to the importance (i.e., the test-over-training density ratio). However, as real-world machine learning tasks are becoming increasingly complex, high-dimensional, and dynamical, novel approaches are explored to cope with such challenges recently. In this article, after introducing the foundation of TL based on importance-weighting, we review recent advances based on joint and dynamic importance-predictor estimation. Furthermore, we introduce a method of causal mechanism transfer that incorporates causal structure in TL. Finally, we discuss future perspectives of TL research.