PCA: Set covariance to zero for empty channels
e-koch opened this issue · comments
At some point, this got fixed. Small non-zero values are placed in the covariance matrix for empty channels, which is needed to ensure the decomposition works and the eigenvalues are real.