Tag
论文提出 SPD-MetaFormer,一种面向小数据脑信号解码的无注意力架构,基于 SPD 流形上均匀加权 Fréchet 聚合,并发现现有流形注意力模型(MAtt、GBWAtt)的学到权重接近均匀,因此简化设计即可获得有竞争力的 EEG 基准结果。
The paper introduces MatFAE, a functional neural network for learning representations from trajectories on the symmetric positive definite (SPD) manifold, with applications in neuroimaging and other scientific domains.
SPDM introduces a geometry-aware state-space model that uses manifold constraints on the symmetric positive definite manifold for time series forecasting, achieving state-of-the-art performance on 11 benchmarks.