Research map / Learning theory and algorithms / Optimal transport
Sliced-Wasserstein on Symmetric Positive Definite Matrices for M/EEG Signals
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use SPDSW-kernel ridge regression on distributions of covariance matrices instead of a single-covariance feature baseline | Cam-CAN brain-age prediction · Cam-CAN | MAE | Helped |
| Use SPDSW rather than logSW for distribution regression | Cam-CAN brain-age prediction · Cam-CAN | MAE | Helped |
| Use distributions of covariance estimates rather than a single average covariance | Cam-CAN brain-age prediction · Cam-CAN | MAE | Helped |
| Use Log-Euclidean kernel ridge regression rather than Ridge regression | Cam-CAN brain-age prediction · Cam-CAN | MAE | Helped |
| Use at least 500 SPDSW projections rather than fewer projections (ablation) | SPDSW-kernel ridge regression · Cam-CAN brain-age prediction | MAE | Helped |
| Vary the covariance timeframe length (ablation) | SPDSW-kernel ridge regression · Cam-CAN brain-age prediction | MAE | Mixed |
| Sample projection directions from symmetric matrices (SPDSW) rather than the ordinary log-space sphere (logSW) (ablation) | Distribution regression on SPD matrices · Cam-CAN brain-age prediction | MAE | Helped |
| Use SPDSW for domain adaptation instead of LEW or LES | BCI domain adaptation · BCI IV 2.a cross-session | runtime (s): 4.34 vs 11.41; 4.34 vs 12.04 | Helped |
| Use SPDSW rather than other transport discrepancies for domain adaptation | BCI domain adaptation · BCI IV 2.a cross-session | accuracy: 82.43; 82.54; 82.24; 82.93 | No clear effect |
| Optimize SPDSW alignment directly over particles rather than using transformations (ablation) | BCI domain adaptation · BCI IV 2.a cross-session | accuracy: 82.43 -> 83.71 | No clear effect |
| Optimize SPDSW alignment directly over particles rather than using transformations (ablation) | BCI domain adaptation · BCI IV 2.a cross-subject | accuracy: 60.63 -> 61.79 | No clear effect |
| Use at least 500 projections for SPDSW alignment (ablation) | BCI domain adaptation · BCI IV 2.a cross-session and cross-subject | accuracy | Helped |
Closest papers by meaning
- Geodesic Optimization for Predictive Shift Adaptation on EEG data (NeurIPS 2024 · similarity 0.87)
- SPDIM: Source-Free Unsupervised Conditional and Label Shift Adaptation in EEG (ICLR 2025 · similarity 0.85)
- Sliced Wasserstein Estimation with Control Variates (ICLR 2024 · similarity 0.84)
- Implicit Gaussian process representation of vector fields over arbitrary latent manifolds (ICLR 2024 · similarity 0.84)
- BrainOmni: A Brain Foundation Model for Unified EEG and MEG Signals (NeurIPS 2025 · similarity 0.83)
- Statistical and Computational Guarantees of Kernel Max-Sliced Wasserstein Distances (ICML 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.