TimeMachine: A Time Series is Worth 4 Mambas for Long-term Forecasting

TL;DR

TimeMachine is a novel time series forecasting model that effectively captures long-term dependencies with linear scalability and low memory usage, outperforming existing methods on benchmark datasets.

Contribution

The paper introduces TimeMachine, a new model utilizing Mamba state-space architecture for scalable, efficient long-term time series forecasting with multi-scale contextual understanding.

Findings

Achieves superior prediction accuracy on benchmarks

Maintains linear scalability and small memory footprint

Outperforms existing models in efficiency and accuracy

Abstract

Long-term time-series forecasting remains challenging due to the difficulty in capturing long-term dependencies, achieving linear scalability, and maintaining computational efficiency. We introduce TimeMachine, an innovative model that leverages Mamba, a state-space model, to capture long-term dependencies in multivariate time series data while maintaining linear scalability and small memory footprints. TimeMachine exploits the unique properties of time series data to produce salient contextual cues at multi-scales and leverage an innovative integrated quadruple-Mamba architecture to unify the handling of channel-mixing and channel-independence situations, thus enabling effective selection of contents for prediction against global and local contexts at different scales. Experimentally, TimeMachine achieves superior performance in prediction accuracy, scalability, and memory efficiency, as extensively validated using benchmark datasets. Code availability: https://github.com/Atik-Ahamed/TimeMachine