ADAPT^2: Adapting Pre-Trained Sensing Models to End-Users via Self-Supervision Replay

arXiv:2404.15305v1 Announce Type: cross
Abstract: Self-supervised learning has emerged as a method for utilizing massive unlabeled data for pre-training models, providing an effective feature extractor for various mobile sensing applications. However, when deployed to end-users, these models encounter significant domain shifts attributed to user diversity. We investigate the performance degradation that occurs when self-supervised models are fine-tuned in heterogeneous domains. To address the issue, we propose ADAPT^2, a few-shot domain adaptation framework for personalizing self-supervised models. ADAPT2 proposes self-supervised meta-learning …

abstract adapt applications arxiv cs.lg data diversity domain eess.sp feature however massive mobile performance pre-training self-supervised learning sensing supervised learning supervision training training models type via