Towards Stable and Structured Time Series Generation with Perturbation-Aware Flow Matching

2025-11-18

AI Summary

Existing time-series generation methods—particularly those based on flow matching—employ globally shared parameters to model the velocity field, limiting their ability to capture spatiotemporal heterogeneity and abrupt transitions induced by local perturbations. This leads to structural inconsistencies and poor stability in generated sequences. To address this, we propose Perturbation-Aware Flow Matching (PAFM), a novel framework featuring: (i) a perturbation-guided training mechanism and a dual-path velocity field that explicitly decouples steady-state evolution from perturbation response; and (ii) a Mixture-of-Experts decoder with flow-guided dynamic routing, enabling perturbation-sensitive trajectory deviation modeling. Evaluated on both unconditional and conditional generation tasks, PAFM consistently outperforms state-of-the-art baselines, delivering significant improvements in structural fidelity, temporal consistency, and robustness against perturbations.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Natural Language Processing: Generation
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Time series generation is critical for a wide range of applications, which greatly supports downstream analytical and decision-making tasks. However, the inherent temporal heterogeneous induced by localized perturbations present significant challenges for generating structurally consistent time series. While flow matching provides a promising paradigm by modeling temporal dynamics through trajectory-level supervision, it fails to adequately capture abrupt transitions in perturbed time series, as the use of globally shared parameters constrains the velocity field to a unified representation. To address these limitations, we introduce extbf{PAFM}, a extbf{P}erturbation- extbf{A}ware extbf{F}low extbf{M}atching framework that models perturbed trajectories to ensure stable and structurally consistent time series generation. The framework incorporates perturbation-guided training to simulate localized disturbances and leverages a dual-path velocity field to capture trajectory deviations under perturbation, enabling refined modeling of perturbed behavior to enhance the structural coherence. In order to further improve sensitivity to trajectory perturbations while enhancing expressiveness, a mixture-of-experts decoder with flow routing dynamically allocates modeling capacity in response to different trajectory dynamics. Extensive experiments on both unconditional and conditional generation tasks demonstrate that PAFM consistently outperforms strong baselines. Code is available at https://anonymous.4open.science/r/PAFM-03B2.

Problem

• Modeling abrupt transitions in perturbed time series data
• Addressing temporal heterogeneity caused by localized perturbations
• Capturing trajectory deviations under perturbation for structural consistency

Innovation

• Perturbation-aware flow matching for time series generation
• Dual-path velocity field capturing trajectory deviations
• Mixture-of-experts decoder with dynamic flow routing