Deciphering Invariant Feature Decoupling in Source-free Time Series Forecasting with Proxy Denoising
It addresses the problem of adapting pretrained models to sparse target domains without source data access, which is incremental as it builds on existing domain adaptation methods.
The paper tackles source-free domain adaptation for time series forecasting by proposing TimePD, a framework that uses large language models and proxy denoising, achieving an average 9.3% improvement over state-of-the-art baselines.
The proliferation of mobile devices generates a massive volume of time series across various domains, where effective time series forecasting enables a variety of real-world applications. This study focuses on a new problem of source-free domain adaptation for time series forecasting. It aims to adapt a pretrained model from sufficient source time series to the sparse target time series domain without access to the source data, embracing data protection regulations. To achieve this, we propose TimePD, the first source-free time series forecasting framework with proxy denoising, where large language models (LLMs) are employed to benefit from their generalization capabilities. Specifically, TimePD consists of three key components: (1) dual-branch invariant disentangled feature learning that enforces representation- and gradient-wise invariance by means of season-trend decomposition; (2) lightweight, parameter-free proxy denoising that dynamically calibrates systematic biases of LLMs; and (3) knowledge distillation that bidirectionally aligns the denoised prediction and the original target prediction. Extensive experiments on real-world datasets offer insight into the effectiveness of the proposed TimePD, outperforming SOTA baselines by 9.3% on average.