Domain Aggregation Networks for Multi-Source Domain Adaptation

TL;DR

This paper introduces Domain Aggregation Networks (DARN), a theoretically grounded method for multi-source domain adaptation that dynamically weights source domains to improve target task performance.

Contribution

It extends domain discrepancy minimization with a finite-sample generalization bound and proposes a new optimization procedure for effective domain weighting.

Findings

DARN outperforms state-of-the-art methods on sentiment analysis datasets.

DARN effectively adjusts source domain weights during training.

Theoretical analysis supports the generalization ability of DARN.

Abstract

In many real-world applications, we want to exploit multiple source datasets of similar tasks to learn a model for a different but related target dataset -- e.g., recognizing characters of a new font using a set of different fonts. While most recent research has considered ad-hoc combination rules to address this problem, we extend previous work on domain discrepancy minimization to develop a finite-sample generalization bound, and accordingly propose a theoretically justified optimization procedure. The algorithm we develop, Domain AggRegation Network (DARN), is able to effectively adjust the weight of each source domain during training to ensure relevant domains are given more importance for adaptation. We evaluate the proposed method on real-world sentiment analysis and digit recognition datasets and show that DARN can significantly outperform the state-of-the-art alternatives.