Closed oesteban closed 4 months ago
wondering what was wrong with the previous approach (besides the standardization of the data, and being a lot less elegant than this one).
The covariance calculation was off. np.cov
requires only one array of shape (n_features, n_samples). Features here is orientations at b=90. The output is a (90, 90) matrix. In your book you get a size in the thousands after the for loop.
Supersedes #215.