Invariant Representation Learning for Source-Free Time Series Forecasting with LLM-Centric Proxy Denoising

2025/10/07 by Keyang Yan, Kangjia Yan, Chenxi Liu +13
Computer Science · Engineering · #Anomaly Detection Techniques and Applications #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Fault Detection and Control Systems #Machine Learning (cs.LG) #Time Series Analysis and Forecasting #cs.AI #cs.LG

paper · pdf · doi:10.48550/arxiv.2510.05589

Accepted by ICML2026

openalex publication_date 2025/10/07 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28 · arxiv created 2026/08/06 · arxiv updated 2026/08/07

Abstract

Effective time series forecasting enables various real-world applications, benefiting from the proliferation of mobile devices. However, the volume of time series data may vary significantly across domains due to high data acquisition costs and data regulations. To maximally create value from sparse data, this study focuses on a new problem of source-free time series forecasting, aiming to adapt a pretrained model from sufficient source time series to the sparse target time series without access to the source data, enabling data protection. To achieve this, we propose TimeID, a novel source-free time series forecasting framework with a large language model (LLM) centric proxy denoising inspired by the powerful generalization capabilities of LLMs. Specifically, TimeID 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 demonstrate that TimeID outperforms state-of-the-art baselines, improving MSE and MAE by 10.7% and 9.3% on average. The code is available at https://github.com/decisionintelligence/TimeID.

Citations

• Moment Alignment: Unifying Gradient and Hessian Matching for Domain Generalization
• BLAST: Balanced Sampling Time Series Corpus for Universal Forecasting Models
• TimeKAN: KAN-based Frequency Decomposition Learning Architecture for Long-term Time Series Forecasting
• WPMixer: Efficient Multi-Resolution Mixing for Long-Term Time Series Forecasting
• Proxy Denoising for Source-Free Domain Adaptation
• TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment
• Hierarchical Classification Auxiliary Network for Time Series Forecasting
• Time-FFM: Towards LM-Empowered Federated Foundation Model for Time Series Forecasting
• TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting
• CALF: Aligning LLMs for Time Series Forecasting via Cross-modal Fine-Tuning
• AirPhyNet: Harnessing Physics-Guided Neural Networks for Air Quality Prediction
• UniTime: A Language-Empowered Unified Model for Cross-Domain Time Series Forecasting
• iTransformer: Inverted Transformers Are Effective for Time Series Forecasting
• Time-LLM: Time Series Forecasting by Reprogramming Large Language Models
• Source-Free Domain Adaptation with Temporal Imputation for Time Series Data
• A Comprehensive Survey on Source-free Domain Adaptation
• A Comprehensive Survey on Source-Free Domain Adaptation
• Source-Free Unsupervised Domain Adaptation: A Survey
• A Time Series is Worth 64 Words: Long-term Forecasting with Transformers
• TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis
• Are Transformers Effective for Time Series Forecasting?
• Towards Spatio-Temporal Aware Traffic Time Series Forecasting--Full Version
• FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
• Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation
• Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting
• Learning explanations that are hard to vary
• Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting
• Domain Adaptation without Source Data
• Universal Source-Free Domain Adaptation
• Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation
• Modeling Long- and Short-Term Temporal Patterns with Deep Neural Networks
• A Survey on Transfer Learning

Related

• CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting
• A Unified Frequency Domain Decomposition Framework for Interpretable and Robust Time Series Forecasting
• Towards a General Time Series Forecasting Model with Unified Representation and Adaptive Transfer
• SDE: A Simplified and Disentangled Dependency Encoding Framework for State Space Models in Time Series Forecasting
• TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting