We are going to implement the CPLE algorithm in Python using a subset extracted from the MNIST dataset. For simplicity, we are going to use only the samples representing the digits 0 and 1:
from sklearn.datasets import load_digitsimport numpy as npX_a, Y_a = load_digits(return_X_y=True)X = np.vstack((X_a[Y_a == 0], X_a[Y_a == 1]))Y = np.vstack((np.expand_dims(Y_a, axis=1)[Y_a==0], np.expand_dims(Y_a, axis=1)[Y_a==1]))nb_samples = X.shape[0]nb_dimensions = X.shape[1]nb_unlabeled = 150Y_true = np.zeros((nb_unlabeled,))unlabeled_idx = np.random.choice(np.arange(0, nb_samples, 1), replace=False, size=nb_unlabeled)Y_true = Y[unlabeled_idx].copy()Y[unlabeled_idx] = -1
After creating the restricted ...
