TimeMachine: A Time Series is Worth 4 Mambas for Long-Term Forecasting
Year: 2024
Type: book-chapter
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.
DOI: 10.3233/faia240677
Authors: Md Atik Ahamed, Qiang Cheng
Institution: University of Kentucky
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Awards: EAGER: PBI: Modeling Place-based Innovation by Leveraging AI-Enabled Dynamic Graph Techniques, Gene Profile Analysis of Associations between Alzheimer’s Disease and Circadian Rhythmic Patterns of Human Brain Regions Using Machine Learning, RI: Small: Optimal Transport Generative Adversarial Networks: Theory, Algorithms, and Applications, Noninvasive Noncontact High-Density Optical Imaging of Neonatal Intraventricular Hemorrhage, G8375851502, University of Kentucky Alzheimer's Disease Research Center, Alzheimer's Disease Research Center
OpenAlex ID: W4403488223