draft: upgrade to numpy 2

epftoolbox/data/_wrangling.py

@@ -17,56 +17,34 @@ def __init__(self):
         self.fitted = False
 
     def fit(self, data):
-
         if len(data.shape)!=2:
-            raise IndexError('Error: Provide 2-D array. First dimension is datapoints and' + 
-                  ' second features')
-            return -1
-
+            raise IndexError('Error: Provide 2-D array. First dimension is datapoints and second features')
         self.median = np.median(data, axis=0)
         self.mad = mad(data, axis=0)
         self.fitted = True
 
     def fit_transform(self, data):
-
         self.fit(data)
         return self.transform(data)
 
     def transform(self, data):
-
         if not self.fitted:
-            print('Error: The scaler has not been yet fitted. Called fit or fit_transform')
-            return -1
+            raise RuntimeError('Error: The scaler has not been yet fitted. Called fit or fit_transform')
+        if len(data.shape) != 2:
+            raise IndexError('Error: Provide 2-D array. First dimension is datapoints and second features')
 
-        if len(data.shape)!=2:
-            raise IndexError('Error: Provide 2-D array. First dimension is datapoints and' + 
-                  ' second features')
-
-        transformed_data = np.zeros(shape=data.shape)
+        numerator = data - self.median
+        denominator = self.mad
+        return np.divide(numerator, denominator, out=np.zeros_like(numerator), where=denominator != 0)
 
-        for i in range(data.shape[1]):
-
-            transformed_data[:, i] = (data[:, i] - self.median[i]) / self.mad[i]
-
-        return transformed_data
 
     def inverse_transform(self, data):
-
         if not self.fitted:
-            print('Error: The scaler has not been yet fitted. Called fit or fit_transform')
-            return -1
-
+            raise RuntimeError('Error: The scaler has not been yet fitted. Call fit or fit_transform')
         if len(data.shape)!=2:
-            raise IndexError('Error: Provide 2-D array. First dimension is datapoints and' + 
-                  ' second features')
-
-        transformed_data = np.zeros(shape=data.shape)
+            raise IndexError('Error: Provide 2-D array. First dimension is datapoints and second features')
 
-        for i in range(data.shape[1]):
-
-            transformed_data[:, i] = data[:, i] * self.mad[i] + self.median[i] 
-
-        return transformed_data
+        return data * self.mad + self.median
 
 
 class InvariantScaler(MedianScaler):
@@ -343,4 +321,4 @@ def scaling(datasets, normalize):
 
         datasets[i] = dataset
 
-    return datasets, scaler
+    return datasets, scaler

epftoolbox/models/_dnn.py

@@ -112,13 +112,13 @@ def __init__(self, neurons, n_features, outputShape=24, dropout=0, batch_normali
             opt = 'adam'
         else:
             if optimizer == 'adam':
-                opt = kr.optimizers.Adam(lr=lr, clipvalue=10000)
+                opt = kr.optimizers.Adam(learning_rate=lr, clipvalue=10000)
             if optimizer == 'RMSprop':
-                opt = kr.optimizers.RMSprop(lr=lr, clipvalue=10000)
+                opt = kr.optimizers.RMSprop(learning_rate=lr, clipvalue=10000)
             if optimizer == 'adagrad':
-                opt = kr.optimizers.Adagrad(lr=lr, clipvalue=10000)
+                opt = kr.optimizers.Adagrad(learning_rate=lr, clipvalue=10000)
             if optimizer == 'adadelta':
-                opt = kr.optimizers.Adadelta(lr=lr, clipvalue=10000)
+                opt = kr.optimizers.Adadelta(learning_rate=lr, clipvalue=10000)
 
         self.model.compile(loss=loss, optimizer=opt)
 
@@ -161,20 +161,19 @@ def _build_model(self):
         for k, neurons in enumerate(self.neurons):
 
             if self.activation == 'LeakyReLU':
-                past_Dense = Dense(neurons, activation='linear', batch_input_shape=inputShape,
+                past_Dense = Dense(neurons, activation='linear', 
                                    kernel_initializer=self.initializer,
                                    kernel_regularizer=self._reg(self.lambda_reg))(past_Dense)
                 past_Dense = LeakyReLU(alpha=.001)(past_Dense)
 
             elif self.activation == 'PReLU':
-                past_Dense = Dense(neurons, activation='linear', batch_input_shape=inputShape,
+                past_Dense = Dense(neurons, activation='linear',
                                    kernel_initializer=self.initializer,
                                    kernel_regularizer=self._reg(self.lambda_reg))(past_Dense)
                 past_Dense = PReLU()(past_Dense)
 
             else:
                 past_Dense = Dense(neurons, activation=self.activation,
-                                   batch_input_shape=inputShape,
                                    kernel_initializer=self.initializer,
                                    kernel_regularizer=self._reg(self.lambda_reg))(past_Dense)
 

epftoolbox/models/_lear.py

@@ -412,7 +412,7 @@ def evaluate_lear_in_test_dataset(path_datasets_folder=os.path.join('.', 'datase
         data_available = pd.concat([df_train, df_test.loc[:date + pd.Timedelta(hours=23), :]], axis=0)
 
         # We set the real prices for current date to NaN in the dataframe of available data
-        data_available.loc[date:date + pd.Timedelta(hours=23), 'Price'] = np.NaN
+        data_available.loc[date:date + pd.Timedelta(hours=23), 'Price'] = np.nan
 
         # Recalibrating the model with the most up-to-date available data and making a prediction
         # for the next day