Leveraging Mixture Alignment for Multi-Source Domain Adaptation

In a conventional Domain Adaptation (DA) setting, we only have one source and target domain, whereas, in many real-world applications, data is often collected from several related sources in different conditions. This has led to a more practical and challenging knowledge transfer problem called Multi-source Domain Adaptation (MDA). Several methodologies, such as prototype matching, explicit distance discrepancy, adversarial learning, etc., have been considered to tackle the MDA problem in recent years. Among them, the adversarial-based learning framework is a popular methodology for transferring knowledge from multiple sources to target domains using a min-max optimization strategy. Despite the advances in adversarial-based methods, several limitations exist, such as the need for a classifier-aware discrepancy metric to align the domains and the need to consider target samples’ consistency and semantic information while aligning the domains. To mitigate these issues, in this work, we propose a novel adversarial learning MDA algorithm, MDAMA, which aligns the target domain with a mixture distribution that consists of source domains. MDAMA uses margin-based discrepancy and augmented intermediate distributions to align the domains effectively. We also propose consistency of target samples by confidence thresholding and transfer of semantic information from multiple source domains to the augmented target domain to further improve the performance of the target domain. We extensively experiment with the MDAMA algorithm on popular real-world MDA datasets such as OfficeHome, Office31, PACS, Office-Caltech, and DomainNet. We evaluate the MDAMA model on these benchmark datasets and demonstrate top performance in all of them.

I. Introduction

Over the last decade, Deep Learning (DL) has achieved tremendous success in various fields, such as agriculture [1], medical imaging [2], robotics [3], etc. Despite this success, traditional DL models underperform when faced with domain shift, i.e., the train and the test set are no longer identically distributed, which is often encountered in the real world [4]. To address this scenario, different methods have been developed over the last few years under the framework of Domain Adaptation (DA) [5]. The main idea behind DA is to adapt the model trained on the labeled source domain to minimize the generalization error on the unlabeled target domain [6]. While conventional DA models typically work with single-source and single-target domains, in many real-world applications, the data is often collected from multiple related domains, such as images taken under different conditions (e.g., lighting, pose, etc.) [7]. This has led to the more practical and challenging Multi-Source Domain Adaptation (MDA) framework, which aims to transfer knowledge from multiple labeled source domains to the unlabeled target domain.