Get.Def.Par.Neural.Network.A3C: Get Default Parameters of the Feed-Forward Neural Network for...
In MartinKies/USLR: Reinforcement Learning with R

Returns a list with default Parameters which could be used as the model.par argument:

name Name of the network. Used to identify the type of network.
setup Function to create the neural network. Default case: Setup.Neural.Network.A3C.
predict Function predicts values based on the neural network. Default case: Predict.Neural.Network.A3C
hidden.nodes Vector containing the number of neurons of each network layer.
activation.hidden Specifies the activation function of the neurons in each hidden layer.
activation.output Specifies the activation functions in the output layer.
c.value Importance of Value loss in the total loss of the network. By Default 0.5.
c.entropy Importance of Entropy loss in the total loss of the network. By Default 0.02.
Optimizer Specifies the used Optimizer and the learning rate. Default Case: Adam Optimizer with a learning rate of 0.005.
decay
clip.norm The global norm of the gradients at which they are clipped. Prevents really large weight updates.