![[2602.22266] WaveSSM: Multiscale State-Space Models for Non-stationary Signal Attention](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.22266] WaveSSM: Multiscale State-Space Models for Non-stationary Signal Attention
Summary
The paper introduces WaveSSM, a novel multiscale state-space model designed to enhance the modeling of non-stationary signals, outperforming existing methods in tasks requiring precise localization.
Why It Matters
WaveSSM addresses limitations of traditional state-space models by utilizing wavelet frames, which provide localized temporal support. This innovation is crucial for applications in fields like biomedical signal processing and audio analysis, where transient dynamics are prevalent.
Key Takeaways
- WaveSSM utilizes wavelet frames for improved signal localization.
- Outperforms traditional polynomial-based state-space models.
- Demonstrated effectiveness on real-world datasets, including physiological signals.
- Addresses challenges in modeling non-stationary signals.
- Provides a foundation for future research in dynamic signal processing.
Computer Science > Machine Learning arXiv:2602.22266 (cs) [Submitted on 25 Feb 2026] Title:WaveSSM: Multiscale State-Space Models for Non-stationary Signal Attention Authors:Ruben Solozabal, Velibor Bojkovic, Hilal Alquabeh, Klea Ziu, Kentaro Inui, Martin Takac View a PDF of the paper titled WaveSSM: Multiscale State-Space Models for Non-stationary Signal Attention, by Ruben Solozabal and 5 other authors View PDF HTML (experimental) Abstract:State-space models (SSMs) have emerged as a powerful foundation for long-range sequence modeling, with the HiPPO framework showing that continuous-time projection operators can be used to derive stable, memory-efficient dynamical systems that encode the past history of the input signal. However, existing projection-based SSMs often rely on polynomial bases with global temporal support, whose inductive biases are poorly matched to signals exhibiting localized or transient structure. In this work, we introduce \emph{WaveSSM}, a collection of SSMs constructed over wavelet frames. Our key observation is that wavelet frames yield a localized support on the temporal dimension, useful for tasks requiring precise localization. Empirically, we show that on equal conditions, \textit{WaveSSM} outperforms orthogonal counterparts as S4 on real-world datasets with transient dynamics, including physiological signals on the PTB-XL dataset and raw audio on Speech Commands. Subjects: Machine Learning (cs.LG); Sound (cs.SD) Cite as: arXiv:2602.22266 [cs....
