ss-Mamba: Semantic-Spline Selective State-Space Model

2025/06/03 by Zuochen Ye, Ye, Zuochen
Computer Science · Decision Sciences · #Artificial Intelligence (cs.AI) #Complex Systems and Decision Making #FOS: Computer and information sciences #Machine Learning (cs.LG) #Neural Networks and Applications #Simulation Techniques and Applications

paper · pdf · doi:10.48550/arxiv.2506.14802

openalex publication_date 2025/06/03 · openalex created_date 2025/10/19 · openalex updated_date 2026/07/28

Abstract

We propose ss-Mamba, a novel foundation model that enhances time series forecasting by integrating semantic-aware embeddings and adaptive spline-based temporal encoding within a selective state-space modeling framework. Building upon the recent success of Transformer architectures, ss-Mamba adopts the Mamba selective state space model as an efficient alternative that achieves comparable performance while significantly reducing computational complexity from quadratic to linear time. Semantic index embeddings, initialized from pretrained language models, allow effective generalization to previously unseen series through meaningful semantic priors. Additionally, spline-based Kolmogorov-Arnold Networks (KAN) dynamically and interpretably capture complex seasonalities and non-stationary temporal effects, providing a powerful enhancement over conventional temporal feature encodings. Extensive experimental evaluations confirm that ss-Mamba delivers superior accuracy, robustness, and interpretability, demonstrating its capability as a versatile and computationally efficient alternative to traditional Transformer-based models in time-series forecasting.

Citations

• ms-Mamba: Multi-scale Mamba for Time-Series Forecasting
• Exploring Kolmogorov-Arnold Networks for Interpretable Time Series Classification
• Kolmogorov-Arnold Networks for Time Series: Bridging Predictive Power and Interpretability
• Intervention-Aware Forecasting: Breaking Historical Limits from a System Perspective
• Empowering Time Series Analysis with Foundation Models: A Comprehensive Survey
• MOMENT: A Family of Open Time-series Foundation Models
• Mamba: Linear-Time Sequence Modeling with Selective State Spaces
• A decoder-only foundation model for time-series forecasting
• Time2Vec: Learning a Vector Representation of Time
• Attention Is All You Need
• On the difficulty of training Recurrent Neural Networks
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Related

• Is Mamba Effective for Time Series Forecasting?
• MambaTS: Improved Selective State Space Models for Long-term Time Series Forecasting
• SST: Multi-Scale Hybrid Mamba-Transformer Experts for Time Series Forecasting
• FMamba: Mamba based on Fast-attention for Multivariate Time-series Forecasting
• Bi-Mamba+: Bidirectional Mamba for Time Series Forecasting