Natural gradients for deep layers

[parvkpr]

In NES algorithms do we use backpropogation from the last layers gradients(computed by the objectve function). I am curious as to how to optimize the hidden layers since they are not directly affecting the objective function

[jinyeom]

Evolution strategies do not use backpropagation, but they differentiate (approximately) across a population of solutions, rather than across parameters within a solution. Perhaps reading this blog post is useful 😛