Online Meta-Learning for Multi-Source and Semi-Supervised Domain Adaptation

TL;DR

This paper introduces an online meta-learning framework to enhance domain adaptation methods, improving performance across multiple datasets and settings by optimizing initial conditions of existing algorithms.

Contribution

It proposes a computationally efficient online meta-learning approach that is agnostic to the base DA algorithm, applicable to multi-source and semi-supervised domain adaptation.

Findings

Achieved state-of-the-art results on several DA benchmarks.

Demonstrated improvements on classic and recent DA techniques.

Validated effectiveness on large-scale DomainNet dataset.

Abstract

Domain adaptation (DA) is the topical problem of adapting models from labelled source datasets so that they perform well on target datasets where only unlabelled or partially labelled data is available. Many methods have been proposed to address this problem through different ways to minimise the domain shift between source and target datasets. In this paper we take an orthogonal perspective and propose a framework to further enhance performance by meta-learning the initial conditions of existing DA algorithms. This is challenging compared to the more widely considered setting of few-shot meta-learning, due to the length of the computation graph involved. Therefore we propose an online shortest-path meta-learning framework that is both computationally tractable and practically effective for improving DA performance. We present variants for both multi-source unsupervised domain adaptation (MSDA), and semi-supervised domain adaptation (SSDA). Importantly, our approach is agnostic to the base adaptation algorithm, and can be applied to improve many techniques. Experimentally, we demonstrate improvements on classic (DANN) and recent (MCD and MME) techniques for MSDA and SSDA, and ultimately achieve state of the art results on several DA benchmarks including the largest scale DomainNet.