You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
hello, wanted to ask u if possible, why did u choose this number of input and output in the layers?
layer_1 = self.layer_manager.create_conv_layer(self.tf_inputs_norm, 32, 8, strides=4, activation_fn='relu',
padding='valid', scope='tf_layer_1')
layer_2 = self.layer_manager.create_conv_layer(layer_1, 64, 4, strides=2, activation_fn='relu', padding='valid',
scope='tf_layer_2')
layer_3 = self.layer_manager.create_conv_layer(layer_2, 64, 3, strides=1, activation_fn='relu', padding='valid',
scope='tf_layer_3')
layer_4 = self.layer_manager.create_fully_connected_layer(layer_3, 512, activation_fn='relu',
scope='tf_layer_4')
as i know the number of outputs in the output layer should be the number of possible actions in the environment, and the input is the state
The text was updated successfully, but these errors were encountered:
The number of hidden layers is based on this paper https://arxiv.org/abs/1803.02965
It suits for Deep Sea Treasure, you can tune the network to suit your problem
hello, wanted to ask u if possible, why did u choose this number of input and output in the layers?
layer_1 = self.layer_manager.create_conv_layer(self.tf_inputs_norm, 32, 8, strides=4, activation_fn='relu',
padding='valid', scope='tf_layer_1')
layer_2 = self.layer_manager.create_conv_layer(layer_1, 64, 4, strides=2, activation_fn='relu', padding='valid',
scope='tf_layer_2')
layer_3 = self.layer_manager.create_conv_layer(layer_2, 64, 3, strides=1, activation_fn='relu', padding='valid',
scope='tf_layer_3')
layer_4 = self.layer_manager.create_fully_connected_layer(layer_3, 512, activation_fn='relu',
scope='tf_layer_4')
as i know the number of outputs in the output layer should be the number of possible actions in the environment, and the input is the state
The text was updated successfully, but these errors were encountered: