RepSPD: Enhancing SPD Manifold Representation in EEGs via Dynamic Graphs
#EEG #geometric deep learning #symmetric positive definite #brain connectivity #neuroscience #artificial intelligence #electroencephalography
📌 Key Takeaways
- Researchers developed RepSPD, a novel geometric deep learning model for EEG analysis
- Current SPD-based methods overlook frequency-specific synchronization and brain region topologies
- RepSPD implements cross-attention on Riemannian manifold and bidirectional alignment strategy
- Experiments show RepSPD outperforms existing methods with better robustness and generalization
📖 Full Retelling
Researchers led by Haohui Jia and including Zheng Chen, Lingwei Zhu, Xu Cao, Yasuko Matsubara, Takashi Matsubara, and Yasushi Sakurai introduced RepSPD, a novel geometric deep learning model for enhancing EEG analysis, through a paper submitted to arXiv on February 26, 2026, to address limitations in current SPD-based methods that overlook frequency-specific synchronization and local topological structures of brain regions. The paper focuses on decoding brain activity from electroencephalography (EEG), which is crucial for neuroscience and clinical applications, particularly highlighting how geometric learning stands out among recent advances in deep learning for EEG due to its theoretical underpinnings on symmetric positive definite (SPD) manifolds that allow revealing structural connectivity analysis in a physics-grounded manner. RepSPD implements a cross-attention mechanism on the Riemannian manifold to modulate geometric attributes of SPD with graph-derived functional connectivity features, and introduces a global bidirectional alignment strategy to reshape tangent-space embeddings, mitigating geometric distortions caused by curvature and thereby enhancing geometric consistency. According to the researchers, extensive experiments demonstrate that their proposed framework significantly outperforms existing EEG representation methods, exhibiting superior robustness and generalization capabilities, potentially advancing our understanding of brain connectivity and improving diagnostic tools for neurological conditions.
🏷️ Themes
neuroscience, artificial intelligence, brain imaging
📚 Related People & Topics
Electroencephalography
Electrophysiological monitoring method to record electrical activity of the brain
Electroencephalography (EEG) is a method to record an electrogram of the spontaneous electrical activity of the brain. The bio signals detected by EEG have been shown to represent the postsynaptic potentials of pyramidal neurons in the neocortex and allocortex. It is typically non-invasive, with the...
Entity Intersection Graph
No entity connections available yet for this article.
Original Source
--> Computer Science > Artificial Intelligence arXiv:2602.22981 [Submitted on 26 Feb 2026] Title: RepSPD: Enhancing SPD Manifold Representation in EEGs via Dynamic Graphs Authors: Haohui Jia , Zheng Chen , Lingwei Zhu , Xu Cao , Yasuko Matsubara , Takashi Matsubara , Yasushi Sakurai View a PDF of the paper titled RepSPD: Enhancing SPD Manifold Representation in EEGs via Dynamic Graphs, by Haohui Jia and 6 other authors View PDF Abstract: Decoding brain activity from electroencephalography is crucial for neuroscience and clinical applications. Among recent advances in deep learning for EEG, geometric learning stands out as its theoretical underpinnings on symmetric positive definite allows revealing structural connectivity analysis in a physics-grounded manner. However, current SPD-based methods focus predominantly on statistical aggregation of EEGs, with frequency-specific synchronization and local topological structures of brain regions neglected. Given this, we propose RepSPD, a novel geometric deep learning -based model. RepSPD implements a cross-attention mechanism on the Riemannian manifold to modulate the geometric attributes of SPD with graph-derived functional connectivity features. On top of this, we introduce a global bidirectional alignment strategy to reshape tangent-space embeddings, mitigating geometric distortions caused by curvature and thereby enhancing geometric consistency. Extensive experiments demonstrate that our proposed framework significantly outperforms existing EEG representation methods, exhibiting superior robustness and generalization capabilities. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2602.22981 [cs.AI] (or arXiv:2602.22981v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2602.22981 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Haohui Jia [ view email ] [v1] Thu, 26 Feb 2026 13:22:19 UTC (6,823 KB) Full-text links: Access Paper: View a PDF of the paper tit...
Read full article at source