def _value_net(self, input_shapes, output_size):
values = MLP(input_shapes=input_shapes,
hidden_layer_sizes=self._hidden_layer_sizes,
output_size=output_size,
activation=self._activation,
output_activation=self._output_activation)
return values
def _policy_net(self, input_shapes, output_size):
raw_actions = MLP(input_shapes=input_shapes,
hidden_layer_sizes=self._hidden_layer_sizes,
output_size=output_size,
activation=self._activation,
output_activation=self._output_activation,
name='{}/GaussianMLPPolicy'.format(self._name))
return raw_actions
def _shift_and_log_scale_diag_net(self, input_shapes, output_size):
shift_and_log_scale_diag_net = MLP(
input_shapes=input_shapes,
hidden_layer_sizes=self._hidden_layer_sizes,
output_size=output_size,
activation=self._activation,
output_activation=self._output_activation,
name='{}/GaussianMLPPolicy'.format(self._name))
return shift_and_log_scale_diag_net
def _shift_and_log_scale_diag_net(self, input_shapes, output_size):
shift_and_log_scale_diag_net = MLP(
input_shapes=input_shapes,
hidden_layer_sizes=self._hidden_layer_sizes,
output_size=output_size,
activation=self._activation,
output_activation=self._output_activation,
name="{}/GaussianMLPPolicy".format(self._name),
kernel_regularizer=tf.keras.regularizers.l2(0.001),
bias_regularizer=tf.keras.regularizers.l2(0.001),
)
return shift_and_log_scale_diag_net