Sensor Geometry as a Flow-Matching Prior for Multi-Channel Brain Signals
Flow-matching models start from an isotropic Gaussian source, the standard choice when the correlation structure of the data is unknown in advance.
Key points
- Existing EEG generative models nonetheless leave the network to learn this from scratch.
- From the sensor coordinates alone, we build a k-nearest-neighbor graph and take a graph-Matérn function of its Laplacian as the source covariance, so the flow starts from spatially coherent patterns rather than channel-independent noise.
- Across eight EEG datasets and four flow-matching methods, the graph-Matérn source lowers the spectral discrepancy between generated and real signals in the five clinical bands (PSD-KL) on most datasets.
- We show that the improvement stems from the spatial eigenvectors of the local graph of sensor positions, since randomizing the eigenvectors while preserving the eigenvalue spectrum eliminates the gain.
Sources (1)
- [1]Sensor Geometry as a Flow-Matching Prior for Multi-Channel Brain SignalsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 01:44 PM
Flow-matching models start from an isotropic Gaussian source, the standard choice when the correlation structure of the data is unknown in advance.
Existing EEG generative models nonetheless leave the network to learn this from scratch.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 6, 2026Compact Robot Policies Need Fine-Grained Visual Representations
- Oct 6, 2026DensiTok: Making Feed-Forward 3D Gaussian Splatting See More Views Than It Is Given
- Oct 6, 2026Variance-Averse $n$-Step Offline Reinforcement Learning for Sparse Long-Horizon Environments
- Oct 6, 2026StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models
- Oct 6, 2026ESP: Energy-Score Policy for One-Step Multimodal Action Generation