[utils] added some debug stuff

bob/pad/base/utils/helper_functions.py

@@ -473,16 +473,30 @@ def prepare_data_for_hyper_param_grid_search(training_features, n_samples):
     # training_features[0] - training features for the REAL class.
     real = convert_and_prepare_features(
         training_features[0])  # output is array
+
     # training_features[1] - training features for the ATTACK class.
     attack = convert_and_prepare_features(
         training_features[1])  # output is array
 
-    # uniformly select subsets of features:
     real_subset = select_uniform_data_subset(real, n_samples)
     attack_subset = select_uniform_data_subset(attack, n_samples)
 
-    # split the data into train and cross-validation:
     real_train, real_cv = split_data_to_train_cv(real_subset)
     attack_train, attack_cv = split_data_to_train_cv(attack_subset)
+    
+    # basically, convert to numpy array and splits real and attacks in two sets of equal size ...
+    # 1st half of examples are for training, 2nd half are for cross-validation
+
+    # WARNING: this is because the provided n_samples is greater that the number of features for each class
+    n_real = np.array(training_features[0]).shape[0]
+    n_attack = np.array(training_features[1]).shape[0]
+    if np.all(np.array(training_features[0][:np.int(n_real/2)]) == real_train):
+      print("real train is equal to the first half of training_features[0]")
+    if np.all(np.array(training_features[1][:np.int(n_attack/2)]) == attack_train):
+      print("attack train is equal to the first half of training_features[1]")
+    if np.all(np.array(training_features[0][np.int(n_real/2):2*np.int(n_real/2)+1]) == real_cv):
+      print("real cv is equal to the second half of training_features[0]")
+    if np.all(np.array(training_features[1][np.int(n_attack/2):2*np.int(n_attack/2)+1]) == attack_cv):
+      print("attack cv is equal to the second half of training_features[1]")
 
     return real_train, real_cv, attack_train, attack_cv