Domain-Generalizable Multiple-Domain Clustering

TL;DR

This paper introduces a novel unsupervised multi-domain clustering approach that learns domain-invariant features and predicts clusters in unseen domains without requiring labeled data or target domain fine-tuning.

Contribution

It proposes a two-stage framework combining self-supervised pre-training and multi-head clustering with pseudo labels and a new label smoothing scheme for domain-generalizable clustering.

Findings

Outperforms baselines requiring target domain fine-tuning

Effective in unseen domains without supervision

Utilizes a novel prediction-based label smoothing scheme

Abstract

This work generalizes the problem of unsupervised domain generalization to the case in which no labeled samples are available (completely unsupervised). We are given unlabeled samples from multiple source domains, and we aim to learn a shared predictor that assigns examples to semantically related clusters. Evaluation is done by predicting cluster assignments in previously unseen domains. Towards this goal, we propose a two-stage training framework: (1) self-supervised pre-training for extracting domain invariant semantic features. (2) multi-head cluster prediction with pseudo labels, which rely on both the feature space and cluster head prediction, further leveraging a novel prediction-based label smoothing scheme. We demonstrate empirically that our model is more accurate than baselines that require fine-tuning using samples from the target domain or some level of supervision. Our code is available at https://github.com/AmitRozner/domain-generalizable-multiple-domain-clustering.