Model Adaptation: Historical Contrastive Learning for Unsupervised Domain Adaptation without Source Data

TL;DR

This paper introduces a novel historical contrastive learning approach for unsupervised domain adaptation that does not require access to source data, effectively aligning target models with source-trained models using historical hypotheses.

Contribution

The paper proposes a new historical contrastive learning technique that enables model adaptation without source data by leveraging historical models for contrastive learning.

Findings

HCL outperforms state-of-the-art methods across various visual tasks.

HCL effectively learns target representations without source data.

The approach maintains source hypothesis integrity during adaptation.

Abstract

Unsupervised domain adaptation aims to align a labeled source domain and an unlabeled target domain, but it requires to access the source data which often raises concerns in data privacy, data portability and data transmission efficiency. We study unsupervised model adaptation (UMA), or called Unsupervised Domain Adaptation without Source Data, an alternative setting that aims to adapt source-trained models towards target distributions without accessing source data. To this end, we design an innovative historical contrastive learning (HCL) technique that exploits historical source hypothesis to make up for the absence of source data in UMA. HCL addresses the UMA challenge from two perspectives. First, it introduces historical contrastive instance discrimination (HCID) that learns from target samples by contrasting their embeddings which are generated by the currently adapted model and the historical models. With the historical models, HCID encourages UMA to learn instance-discriminative target representations while preserving the source hypothesis. Second, it introduces historical contrastive category discrimination (HCCD) that pseudo-labels target samples to learn category-discriminative target representations. Specifically, HCCD re-weights pseudo labels according to their prediction consistency across the current and historical models. Extensive experiments show that HCL outperforms and state-of-the-art methods consistently across a variety of visual tasks and setups.