Simpler brain decoding method works well with limited neural data
SPD-MetaFormer is what you need for small-data brain decoding
Machine Learning
Summary
Brain signals are hard to interpret because they are noisy and vary between people, and there often isn't much labeled data available. The authors studied existing methods that use a technique called attention on special brain signal representations but found that the attention doesn't really focus more on some parts than others. Based on this, they created a simpler method called SPD-MetaFormer that avoids attention and uses uniform weighting instead. Their approach performs competitively in decoding brain signals from EEG data despite using less complex attention mechanisms. This shows that simpler designs can still be effective when data and sequences are short.
What this means in practice
- •For brain-computer interface developers: Design decoding algorithms that use fewer parameters to interpret neural signals from limited EEG data effectively.
- •For medical device engineers: Build simpler EEG-based devices for neurological monitoring that require less complex models but maintain decoding accuracy.
Authors
Zhida Wang, Wei Lyu, Guo Yu, Sui Tang
Abstract
Brain signal decoding is challenging because neural recordings are noisy and vary across individuals, while labeled data are often limited. Recent attention-based models on the symmetric positive definite (SPD) manifold have nevertheless achieved strong performance using covariance and connectivity representations, yet the contribution of learned token weighting remains unclear. We examine two representative architectures, MAtt (based on log-Euclidean geometry) and GBWAtt (based on generalized Bures--Wasserstein geometry), and find that their learned attention weights remain close to uniform after training. We relate this behavior to bounded similarity parameterizations that, under the original softmax scaling, limit attention-weight contrast. Moreover, replacing learned weights with uniform weights, throughout training and evaluation, has little effect on mean predictive performance while preserving each model's original aggregation geometry. Motivated by these findings, we introduce SPD-MetaFormer, an attention-free architecture built on uniformly weighted Fréchet aggregation under log-Euclidean geometry. Its backbone uses a geodesic residual to update a summary token and a shared spectral feedforward map to transform all tokens, followed by a learned weighted readout. Token states remain SPD-valued until tangent-space classification. Across three EEG benchmarks, SPD-MetaFormer achieves competitive results relative to published Euclidean and manifold baselines. Separate matched reproductions test learned versus uniform weighting within MAtt and GBWAtt. These results suggest that, in the short-sequence and limited-data regimes studied, carefully designed SPD architectures can provide a simpler and effective alternative to adaptive manifold attention.