matlab白化代码-csp_evaluation:评估不具有完整等级的协方差矩阵的常见空间模式实现

# Evaluation of common spatial patterns implementations for covariance matrices without full rank ### Content - `implementations`: CSP implementations - Our implementations: - `csp_python.py`: all Python implementations - `csp_geometric_approach_dim_reduction.m`: Matlab geometric approach with dimensionality reduction during the whitening step - `csp_geometric_approach_no_checks.m`: Matlab geometric approach without dimensionality reduction (no checks) - `csp_gep_no_checks.m`: Matlab generalized eigenvalue problem approach (no checks) - Toolboxes: - `use_bbci.m`: BBCI Toolbox - `use_fieldtrip.m`: FieldTrip - `use_biosig.m`: BioSig - `use_mne.py`: MNE - `artifacts_removal`: artifact removal methods based on ICA - `evaluation.py`: Evaluate all CSP implementations on a binary motor imagery classification task ## Replication Create the Conda environment from the `environment.yml` file: ```console conda env create -f environment.yml ``` Activate the new environment: ```console conda activate csp_evaluation ``` Note that CSP implementations from Matlab Toolboxes for EEG analysis contain hard-coded paths for toolboxes locations in their corresponding .m files. Directory `results` already contains pre-computed ICA for each patient. Evaluation can be run in parallel (on the same or multiple computers), which is useful for Matlab implementations, by uncommenting line 166 in `evaluation.py` ```python # @TaskGenerator ``` Then, tasks will be executed using [jug](https://jug.readthedocs.io/en/latest/) package. Run it by ```console jug execute evaluation.py ``` and see the progress by ```console jug status evaluation.py ``` Results can be inspected and visualized by scripts in `visualize` directory.