MPNet: A Robust and Efficient Manifold Pooling Network for Multi-Rhythm EEG Signal Decoding

Guoqing Cai; Kai Zeng; Shoulin Huang; Ting Ma

eess.SPApr 11, 2026

MPNet: A Robust and Efficient Manifold Pooling Network for Multi-Rhythm EEG Signal Decoding

Authors: Guoqing Cai, Kai Zeng, Shoulin Huang, Ting Ma

Organizations: Harbin Institute of Technology (Shenzhen), Shenzhen, China · Guangxi Normal University, Guilin, China · Peng Cheng Laboratory, Shenzhen, China

Abstract

Deep Riemannian networks provide a powerful framework for Electroencephalography (EEG) decoding, but their practical applications are severely constrained. Accurately decoding EEG signals requires modeling complex temporal dynamics across multiple rhythms, which results in high-dimensional Riemannian inputs and significant computational costs. To address this, we propose the Manifold Pooling Network (MPNet). MPNet uses a rhythm-adaptive convolutional frontend to extract comprehensive time-frequency representations and generate multi-view Riemannian nodes. A novel manifold node pooling layer is then proposed to aggregate these nodes into a single fusion node with a fixed size, enabling the following deep Riemannian network to process it with greatly reduced costs. Experiments on two public EEG datasets show that MPNet achieves state-of-the-art accuracy, runs up to 10 times faster than the comparable Riemannian model, and maintains robust performance under limited-data conditions. These findings highlight MPNet's practicality and efficiency for real-world EEG applications.

Figures & tables

Figure 1: The illustration of the proposed MPNet, which comprises three main stages: (a) a Rhythm-Adaptive Convolutional Module , (b) a Manifold Node Pooling Module , and (c) a Riemannian Representation Learning Module

	BCI-IV 2a Dataset			OpenBMI Dataset
Methods	ACC (%)	Kappa (%)	F1 (%)	ACC (%)	Kappa (%)	F1 (%)
EM	80.71	74.28	80.37	79.60	59.20	79.33
WEM	80.40	73.86	80.10	79.10	58.20	78.78
RM	80.03	73.37	79.77	78.77	57.54	78.51
WRM	80.97	74.62	80.74	78.96	57.92	78.47
w/o MP	76.35	68.47	74.68	76.05	51.20	75.81

Table 1: Ablation Study of Manifold Node Pooling Strategies on the BCI-IV 2a and OpenBMI Dataset. The best result is highlighted in bold, and the second-best is underlined.

	BCI-IV 2a Dataset				OpenBMI Dataset
Methods	ACC (%)	Kappa (%)	F1 (%)	$p$ -value	ACC (%)	Kappa (%)	F1 (%)	$p$ -value
EEGNet (2018)	70.52 $\pm$ 15.25	60.70 $\pm$ 20.33	70.11 $\pm$ 15.51	0.0039 ∗∗	70.75 $\pm$ 14.44	41.50 $\pm$ 19.25	70.37 $\pm$ 14.75	$<0.0001$ ∗∗
ShallowNet (2017)	72.34 $\pm$ 11.96	63.12 $\pm$ 15.95	71.84 $\pm$ 12.31	0.0020 ∗∗	74.99 $\pm$ 13.13	49.98 $\pm$ 17.51	74.64 $\pm$ 13.49	$<0.0001$ ∗∗
FBCNet (2021)	77.53 $\pm$ 11.82	70.04 $\pm$ 15.76	77.01 $\pm$ 12.28	0.0098 ∗∗	74.04 $\pm$ 13.98	48.08 $\pm$ 18.64	73.66 $\pm$ 14.33	$<0.0001$ ∗∗
IFNet (2023)	78.24 $\pm$ 10.63	70.98 $\pm$ 14.17	78.01 $\pm$ 10.76	0.0273 ∗	78.13 $\pm$ 13.50	56.26 $\pm$ 18.00	77.95 $\pm$ 13.89	0.0120 ∗
EEGConformer (2023)	79.09 $\pm$ 11.48	72.12 $\pm$ 15.31	78.70 $\pm$ 11.97	0.0371 ∗	78.57 $\pm$ 13.41	57.14 $\pm$ 17.88	78.31 $\pm$ 13.67	0.0205 ∗

Table 2: Comprehensive performance comparison (Mean

\pm

Std) on the BCI-IV 2a and OpenBMI datasets. The best result is highlighted in bold, and the second-best is underlined.

Figure 2: Accuracy and stability across varying training data proportions on the two datasets.

Figure 3: Computational cost comparison. All models were evaluated on an NVIDIA RTX 2080 Ti GPU with 12 GB of memory.

Riemannian Neural Networks

See all 10 papers

Towards Robust EEG Decoding Based on Riemannian Self-Attention
Jun 24, 2026Shaocheng Jin, Tao Zhou, Rui Wang +4Brain-Computer Interfaces Self-Attention
School of Artificial Intelligence and Computer Science Jiangnan University Wuxi, China · Department of Information Engineering and Computer Science University of Trento Trento, Italy · Centre for Vision, Speech and Signal Processing (CVSSP) University of Surrey Guildford, UK
DSAINet: An Efficient Dual-Scale Attentive Interaction Network for General EEG Decoding
Apr 20, 2026Zhiyuan Ma, Zeyuan Li, Zihao Qiu +6Cross-Subject EEG Decoding EEG Decoding
School of Biomedical Engineering, Tsinghua Laboratory of Brain and Intelligence, and IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing 100084, China · School of Basic Medical Sciences and Tsinghua Laboratory of Brain and Intelligence, Tsinghua University, Beijing 100084, China · Department of Psychological and Cognitive Sciences, Tsinghua University, Beijing 100084, China +1
I\textsuperscript{2}RiMA: Spectral Riemannian Representation with Temporal Attention for Mental Stress Detection based on EEG Signals
Jul 1, 2026Cheng He, Kunyu Peng, Shangen Han +3Cross-Subject EEG Decoding Electroencephalography
Laboratory of Neural Computing and Intelligent Perception, College of Information Engineering, Capital Normal University, Beijing, China · Karlsruhe Institute of Technology, Germany · School of Psychology, Capital Normal University, Beijing, China

Jun 24, 2026J/K move · Enter open · S save

Towards Robust EEG Decoding Based on Riemannian Self-Attention

Shaocheng Jin, Tao Zhou, Rui Wang, Ziheng Chen+3

School of Artificial Intelligence and Computer Science Jiangnan University Wuxi, China · Department of Information Engineering and Computer Science University of Trento Trento, Italy · Centre for Vision, Speech and Signal Processing (CVSSP) University of Surrey Guildford, UK

Brain-Computer Interfaces Self-Attention EEG Decoding

Open paper AskPDF arXiv

Brain-Computer Interface (BCI) based on electroencephalography (EEG) enables direct interaction between the brain and external environments and has significant applications in assistive technologies, medical rehabilitation, and entertainment. Recently, EEG decoding methods based on Symmetric Positive Definite (SPD) learning have demonstrated superior performance. However, these methods typically employ basic network architectures and do not explicitly capture local relationships between EEG signals. This limitation is problematic for EEG signals due to their inherently low Signal-to-Noise Ratio (SNR). Moreover, most existing Riemannian manifold-based methods are restricted to specific metrics. The most widely used is the Affine-Invariant Metric (AIM). However, it has a quadratic dependency on the SPD matrices and cannot handle ill-conditioned SPD matrices, which hinders the effectiveness of networks. In contrast, the Bures-Wasserstein Metric (BWM) exhibits linear dependence on SPD matrices and demonstrates superior performance for ill conditioning. To overcome these challenges, we propose a Riemannian self-attention network based on the BWM. Additionally, the recently introduced power-deformed generalized Bures-Wasserstein metric reveals a nonlinear relationship between SPD matrices and matrix power deformation. This metric provides a more nuanced representation of the geometric structure of the SPD manifold. Consequently, we extend our model to a learnable version. For simplicity, we refer to it as GBWAtt. Experimental results on three EEG benchmarking datasets validate the robustness and effectiveness of our proposed method. The code is available at https://github.com/jissc/GBWAtt.

Abstract

Riemannian Neural Networks

Explore similar work

Towards Robust EEG Decoding Based on Riemannian Self-Attention

DSAINet: An Efficient Dual-Scale Attentive Interaction Network for General EEG Decoding

I\textsuperscript{2}RiMA: Spectral Riemannian Representation with Temporal Attention for Mental Stress Detection based on EEG Signals