Network last layer activation function

[anugrahasinha]

The final later of the network should not have an activation of softmax, ideally it should be "linear" activation as we will be measuring the action values, and not the probability of taking an action.

model.add(layers.Dense(self.num_outputs,activation='softmax'))

[johnhelt]

Ok, thanks, I will give it a check.

[anugrahasinha]

I am evaluating the code and trying to tweak a couple of things. Would it possible for you to add me as a contributor to your repo. I will create a develop branch and work in that branch.

[johnhelt]

I've added you as a collaborator. Thanks alot for your kind help!