Tackling unsupervised multi-source domain adaptation with optimism and consistency

TL;DR

This paper introduces a novel framework for unsupervised multi-source domain adaptation that optimizes mixture weights and employs consistency regularization, outperforming existing methods.

Contribution

It proposes a new approach combining optimism and consistency regularization to better handle mixture weights and improve target domain performance.

Findings

Outperforms current state-of-the-art methods.

Effectively adjusts mixture distribution weights.

Reduces target domain error more than existing techniques.

Abstract

It has been known for a while that the problem of multi-source domain adaptation can be regarded as a single source domain adaptation task where the source domain corresponds to a mixture of the original source domains. Nonetheless, how to adjust the mixture distribution weights remains an open question. Moreover, most existing work on this topic focuses only on minimizing the error on the source domains and achieving domain-invariant representations, which is insufficient to ensure low error on the target domain. In this work, we present a novel framework that addresses both problems and beats the current state of the art by using a mildly optimistic objective function and consistency regularization on the target samples.