how to subtract the mean value of the network output values

Hello,

I'm trying to implement Action Branching Architectures for Deep Reinforcement Learning proposed by https://arxiv.org/abs/1711.08946 and since it is my first time training a machine learning on matlab I having a lot of problems. my question is how to subtract the mean value of the network output values from each output? the architecture is shown in the image. I usually do it in this way using Pytorch

advantage_Net = advantage_Net -advantage_Net.mean(1, keepdim=True)

Answer 1

[Chendi Lin]

To normalize the net I think you can simply do

advantage_Net = advantage_Net - mean(advantage_Net)

Please let me know if I understand your question correctly.

CD

[Izzat Brgli]

I will try it, thanks a lot for helping