LFP-Riemannian

Riemannian geometry-based classification of brain states from invasive LFP recordings (Minimum Distance to Mean on covariance matrices).

Where to start:

• Notebook: notebooks/run_riemann.ipynb (single Riemann-only runner with inline constants/helpers)
• Script: main.py (minimal example; update file paths before running)

Features

• Accuracy: Outperforms CNNs and Euclidean MDM classifiers.
• Efficiency: Up to 100x faster training times with smaller datasets.
• Insights: Reveals spatial organization of brain regions across different brain states.

Installation

Clone the repository and install dependencies:

git clone https://github.com/your_username/LFP-Riemannian.git
cd LFP-Riemannian
pip install -r requirements.txt

Usage

• Open notebooks/run_riemann.ipynb and set MODE ("regions" or "train_sizes") and your data file paths in LIST_SUBJECTS.
• Or run main.py after updating its example file paths.

Citation

If you found this useful, consider citing!

Marin-Llobet, A., et al. (2025). Riemannian Geometry for the classification of brain states with intracortical brain-computer interfaces. arXiv preprint arXiv:2504.05534. https://arxiv.org/abs/2504.05534