2
$\begingroup$

I've been recently looking into gradient free learning of neural networks. However, most of the techniques I've found seem to be only applied to toy problems, which I assume means they're infeasible for more "demanding" scenarios, such as MNIST digit classification. Also, I've been unable to find survey papers or sites on the subject.

Is there a good resource to learn about gradient free learning and seeing what SOTA is capable of?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

I believe Deep Neuroevolution is close to state of the art in terms of gradient free methods. They use it for training agents which play Atari games from image inputs.

Anyway, there is not much research activity in training neural networks with GA or other gradient free methods -- because we do have gradients, and they make it much easier! (Of course the same is not completely true in RL.)

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.