AdaMamba: Adaptive Frequency-Gated Mamba for Long-Term Time Series Forecasting

2026-04-25

AI Summary

This work addresses the limitations of existing long-term time series forecasting methods, which often rely on homogeneous assumptions in the time–frequency domain and struggle to effectively model dynamic periodicity and long-range dependencies. To overcome these challenges, the authors propose an adaptive frequency-domain gated state space model that deeply integrates the Mamba architecture with spectral analysis. The approach employs interactive chunk-wise encoding to capture dynamic inter-variable relationships and introduces an input-dependent frequency-domain gating mechanism. This mechanism generalizes the conventional temporal forgetting gate into a unified time–frequency joint forgetting gate, enabling dynamic state calibration based on spectral importance. The method breaks free from cross-domain homogeneity constraints and achieves state-of-the-art performance across seven public long-term forecasting benchmarks and two domain-specific datasets, while maintaining both computational efficiency and high prediction accuracy.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Accurate long-term time series forecasting (LTSF) requires the capture of complex long-range dependencies and dynamic periodic patterns. Recent advances in frequency-domain analysis offer a global perspective for uncovering temporal characteristics. However, real-world time series often exhibit pronounced cross-domain heterogeneity where variables that appear synchronized in the time domain can differ substantially in the frequency domain. Existing frequency-based LTSF methods often rely on implicit assumptions of cross-domain homogeneity, which limits their ability to adapt to such intricate variability. To effectively integrate frequency-domain analysis with temporal dependency learning, we propose AdaMamba, a novel framework that endogenizes adaptive and context-aware frequency analysis within the Mamba state-space update process. Specifically, AdaMamba introduces an interactive patch encoding module to capture inter-variable interaction dynamics. Then, we develop an adaptive frequency-gated state-space module that generates input-dependent frequency bases, and generalizes the conventional temporal forgetting gate into a unified time-frequency forgetting gate. This allows dynamic calibration of state transitions based on learned frequency-domain importance, while preserving Mamba's capability in modeling long-range dependencies. Extensive experiments on seven public LTSF benchmarks and two domain-specific datasets demonstrate that AdaMamba consistently outperforms state-of-the-art methods in forecasting accu racy while maintaining competitive computational efficiency. The code of AdaMamba is available at https://github.com/XDjiang25/AdaMamba.

Problem

• long-term time series forecasting
• frequency-domain heterogeneity
• cross-domain variability
• temporal dependency
• dynamic periodic patterns

Innovation

• Adaptive Frequency Gating
• State-Space Model
• Time-Frequency Forgetting Gate
• Long-Term Time Series Forecasting
• Cross-Domain Heterogeneity