The algorithm is covered in depth on our overleaf page
This connects the variables in the overleaf page to variables used in python. Everything is 0-based indexed and classes are labeled 0,1,2,3...num_classes
-
X stays the same with the same shape
-
Theta -> theta_params. Same shape
-
k -> class_assignments
-
pi -> class_probabilities
-
squiggly pi -> class_membership_scores
-
n_c -> num_members_per_class
-
n_cj -> num_correct_per_class_by_q
-
[alpha,beta] -> theta_hyperparams
-
[alpha_c] -> class_prob_hyperparams