def __init__(self, l1=0., l2=0., **kwargs):
super(ActivityRegularization,
self).__init__(activity_regularizer=regularizers.L1L2(l1=l1,
l2=l2),
**kwargs)
self.supports_masking = True
self.l1 = l1
self.l2 = l2
def kernel(self, train_columns_count):
regression = tf.keras.Sequential([
tf.keras.layers.Dense(self.dense_num_list[0], activation=self.ACTIVATION,
input_shape=[train_columns_count, ]),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(self.dense_num_list[1],
activation=self.ACTIVATION,
regularizers=regularizers.L1L2(0.1, 0.1)),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(1, activation="linear")
])
optimizer = self.get_optimizer()
regression.compile(loss=self.loss_class(), optimizer=optimizer, metrics=['mae', 'mse'])
return regression
