OATS: Online Data Augmentation for Time Series Foundation Models

2026-01-26

AI Summary

This work proposes OATS, an online adaptive data augmentation framework tailored for time series foundation models, addressing the limitations of existing static heuristic strategies that fail to adapt to training dynamics. OATS introduces, for the first time in this domain, the concept of dynamic data optimization by leveraging a diffusion model to generate high-quality synthetic data. It employs a guided exploration–exploitation mechanism that utilizes high-value samples observed during training as steering signals, effectively balancing generation efficiency and augmentation efficacy. Extensive experiments across six diverse datasets and two mainstream foundation models demonstrate that OATS consistently outperforms both standard training and static augmentation baselines, yielding significant performance gains.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Diffusion Models for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Time Series Foundation Models (TSFMs) are a powerful paradigm for time series analysis and are often enhanced by synthetic data augmentation to improve the training data quality. Existing augmentation methods, however, typically rely on heuristics and static paradigms. Motivated by dynamic data optimization, which shows that the contribution of samples varies across training stages, we propose OATS (Online Data Augmentation for Time Series Foundation Models), a principled strategy that generates synthetic data tailored to different training steps. OATS leverages valuable training samples as principled guiding signals and dynamically generates high-quality synthetic data conditioned on them. We further design a diffusion-based framework to produce realistic time series and introduce an explore-exploit mechanism to balance efficiency and effectiveness. Experiments on TSFMs demonstrate that OATS consistently outperforms regular training and yields substantial performance gains over static data augmentation baselines across six validation datasets and two TSFM architectures. The code is available at the link https://github.com/microsoft/TimeCraft.

Problem

• Time Series Foundation Models
• Data Augmentation
• Online Augmentation
• Dynamic Data Optimization
• Synthetic Data

Innovation

• online data augmentation
• time series foundation models
• diffusion-based generation
• dynamic data optimization
• explore-exploit mechanism