LA.eig(sigma) gives complex eigenvalues for symmetrical sigma

[tamkaho]

In line 293 of contrastive/init.py:
w, v = LA.eig(sigma)

It seems that np.linalg.eig sometimes gives complex eigenvalues due to truncation error even though my sigmas are symmetrical matrices. Would it be better to replace the line with the following?

w, v = np.linalg.eigh(sigma)