FlowState: Sampling Rate Invariant Time Series Forecasting

TL;DR

FlowState is a novel time series foundation model that uses a state space model and functional basis decoder to achieve continuous-time modeling, scale adaptability, and improved efficiency, outperforming existing models.

Contribution

Introduction of FlowState, a TSFM with a state space encoder and functional basis decoder for scale-invariant, continuous-time forecasting, with an efficient pretraining strategy.

Findings

FlowState outperforms all other models on GIFT-ZS and Chronos-ZS benchmarks.

FlowState is the smallest model yet achieves state-of-the-art results.

FlowState can adapt online to varying input sampling rates.

Abstract

Foundation models (FMs) have transformed natural language processing, but their success has not yet translated to time series forecasting. Existing time series foundation models (TSFMs), often based on transformer variants, struggle with generalization across varying context and target lengths, lack adaptability to different sampling rates, and are computationally inefficient. We introduce FlowState, a novel TSFM architecture that addresses these challenges through two key innovations: a state space model (SSM) based encoder and a functional basis decoder. This design enables continuous-time modeling and dynamic time-scale adjustment, allowing FlowState to inherently generalize across all possible temporal resolutions, and dynamically adjust the forecasting horizons. In contrast to other state-of-the-art TSFMs, which require training data across all possible sampling rates to memorize patterns at each scale, FlowState inherently adapts its internal dynamics to the input scale, enabling smaller models, reduced data requirements, and improved efficiency. We further propose an efficient pretraining strategy that improves robustness and accelerates training. Despite being the smallest model, FlowState outperforms all other models and is state-of-the-art for the GIFT-ZS and the Chronos-ZS benchmarks. Ablation studies confirm the effectiveness of its components, and we demonstrate its unique ability to adapt online to varying input sampling rates.