Efficient Source-Free Time-Series Adaptation via Parameter Subspace Disentanglement

TL;DR

This paper introduces a parameter-efficient source-free domain adaptation method for time-series data that reparameterizes models using Tucker decomposition, significantly reducing training and inference costs while maintaining performance.

Contribution

It proposes a novel Tucker-style reparameterization for source models, enabling selective fine-tuning and improved efficiency in source-free time-series adaptation.

Findings

Reduces fine-tuned parameters and MACs by over 90%.

Maintains competitive performance with significantly less computation.

Compatible with various SFDA methods.

Abstract

In this paper, we propose a framework for efficient Source-Free Domain Adaptation (SFDA) in the context of time-series, focusing on enhancing both parameter efficiency and data-sample utilization. Our approach introduces an improved paradigm for source-model preparation and target-side adaptation, aiming to enhance training efficiency during target adaptation. Specifically, we reparameterize the source model's weights in a Tucker-style decomposed manner, factorizing the model into a compact form during the source model preparation phase. During target-side adaptation, only a subset of these decomposed factors is fine-tuned, leading to significant improvements in training efficiency. We demonstrate using PAC Bayesian analysis that this selective fine-tuning strategy implicitly regularizes the adaptation process by constraining the model's learning capacity. Furthermore, this re-parameterization reduces the overall model size and enhances inference efficiency, making the approach particularly well suited for resource-constrained devices. Additionally, we demonstrate that our framework is compatible with various SFDA methods and achieves significant computational efficiency, reducing the number of fine-tuned parameters and inference overhead in terms of MACs by over 90% while maintaining model performance.