sort eigen vector in PrincipalComponentAnalysis

[raytroop]

mlxtend/mlxtend/feature_extraction/principal_component_analysis.py

Lines 143 to 153 in a4bcab8

	def _decomposition(self, mat, n_samples):
	if self.solver == 'eigen':
	e_vals, e_vecs = np.linalg.eig(mat)
	elif self.solver == 'svd':
	u, s, v = np.linalg.svd(mat.T)
	e_vecs, e_vals = u, s
	e_vals = e_vals ** 2 / n_samples
	sort_idx = np.argsort(e_vals)[::-1]
	e_vals, e_vecs = e_vals[sort_idx], e_vecs[sort_idx]
	return e_vals, e_vecs

Because the column v[:,i] is the eigenvector,

mlxtend/mlxtend/feature_extraction/principal_component_analysis.py

Line 152 in a4bcab8

e_vals, e_vecs = e_vals[sort_idx], e_vecs[sort_idx]

should be

e_vals, e_vecs = e_vals[sort_idx], e_vecs[:, sort_idx]

[rasbt]

Thanks! Looks like there eigenvectors were already sorted in all the unit tests which is why this wasn't caught in the unit tests. I think this is guaranteed in SVD but not necessarily in np.eig. Will fix it, many thanks!