FeDaL: Federated Dataset Learning for Time Series Foundation Models

2025-08-05

AI Summary

Time series foundation models (TSFMs) suffer from domain bias induced by cross-dataset heterogeneity, severely limiting their generalization capability. To address this, we propose FeDaL—the first federated dataset learning framework for TSFMs—which decouples shared generalizable knowledge from client-specific personalized knowledge via a distributed architecture. FeDaL innovatively introduces dual bias elimination mechanisms: Domain Bias Elimination (to mitigate local client-level biases) and Global Bias Elimination (to correct systemic dataset-level biases), operating synergistically. The framework supports multi-task time-series analysis and incorporates a cross-dataset evaluation protocol. Extensive experiments across eight real-world tasks demonstrate that FeDaL significantly outperforms 54 baseline methods, establishing state-of-the-art cross-domain generalization. Furthermore, we systematically analyze the impact of data volume, number of clients, and client participation rate on decentralized modeling performance.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Multiagent Systems: Adversarial Agents

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Decentralized Web and Fediverse systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Dataset-wise heterogeneity introduces significant domain biases that fundamentally degrade generalization on Time Series Foundation Models (TSFMs), yet this challenge remains underexplored. This paper rethink the development of TSFMs using the paradigm of federated learning. We propose a novel Federated Dataset Learning (FeDaL) approach to tackle heterogeneous time series by learning dataset-agnostic temporal representations. Specifically, the distributed architecture of federated learning is a nature solution to decompose heterogeneous TS datasets into shared generalized knowledge and preserved personalized knowledge. Moreover, based on the TSFM architecture, FeDaL explicitly mitigates both local and global biases by adding two complementary mechanisms: Domain Bias Elimination (DBE) and Global Bias Elimination (GBE). FeDaL`s cross-dataset generalization has been extensively evaluated in real-world datasets spanning eight tasks, including both representation learning and downstream time series analysis, against 54 baselines. We further analyze federated scaling behavior, showing how data volume, client count, and join rate affect model performance under decentralization.

Problem

• Address dataset-wise heterogeneity degrading Time Series Foundation Models
• Propose federated learning to handle heterogeneous time series data
• Mitigate local and global biases via Domain and Global Bias Elimination

Innovation

• Federated learning for time series datasets
• Domain and Global Bias Elimination mechanisms
• Dataset-agnostic temporal representations learning