Implemented shuffle parameter in preprocessing; shuffled validation data in model_training.py

model_training.py

@@ -294,6 +294,7 @@ input_dict_val = {
         random_state=42,
         filter="gaussian",
         sigma=55,
+        shuffle=True,
     ),
     "dists": distances2.preprocess(
         window_size=11,
@@ -302,6 +303,7 @@ input_dict_val = {
         random_state=42,
         filter="gaussian",
         sigma=55,
+        shuffle=True,
     ),
     "angles": angles2.preprocess(
         window_size=11,
@@ -310,6 +312,7 @@ input_dict_val = {
         random_state=42,
         filter="gaussian",
         sigma=55,
+        shuffle=True,
     ),
     "coords+dist": coords_distances2.preprocess(
         window_size=11,
@@ -318,6 +321,7 @@ input_dict_val = {
         random_state=42,
         filter="gaussian",
         sigma=55,
+        shuffle=True,
     ),
     "coords+angle": coords_angles2.preprocess(
         window_size=11,
@@ -326,6 +330,7 @@ input_dict_val = {
         random_state=42,
         filter="gaussian",
         sigma=55,
+        shuffle=True,
     ),
     "coords+dist+angle": coords_dist_angles2.preprocess(
         window_size=11,
@@ -334,6 +339,7 @@ input_dict_val = {
         random_state=42,
         filter="gaussian",
         sigma=55,
+        shuffle=True,
     ),
 }
 

source/preprocess.py

@@ -472,6 +472,7 @@ class table_dict(dict):
         sigma=None,
         shift=0,
         standard_scaler=True,
+        shuffle=False,
     ):
         """Builds a sliding window. If desired, splits train and test and
            Z-scores the data using sklearn's standard scaler"""
@@ -532,8 +533,15 @@ class table_dict(dict):
             if filter == "gaussian":
                 X_test = X_test * g.reshape(1, window_size, 1)
 
+            if shuffle:
+                X_train = np.random.choice(X_train, X_train.shape[0], replace=False)
+                X_test = np.random.choice(X_test, X_test.shape[0], replace=False)
+
             return X_train, X_test
 
+        if shuffle:
+            X_train = np.random.choice(X_train, X_train.shape[0], replace=False)
+
         return X_train
 
     def random_projection(self, n_components=None, sample=1000):