1
$\begingroup$

I am familiar with the basics (and perhaps a substantial amount of basics) of imitation learning and reinforcement learning. In IL (imitation), we take demonstrations from an assumed expert, which we also assume has the most efficient policy.

What does this statement mean: This will train a policy network using the expert dataset and store the results.

Question: Why do I need to TRAIN a policy network? What is it for? Are there any algorithms that come with this training a policy?

Thanks

$\endgroup$
1
$\begingroup$

A policy is just a response or action given a state (situation). Training a neural network with samples from an expert is just searching for a function F that efficiently maps states to actions.

For example for chess, a neural network can be trained with the action to be taken in a given state of the board. This can be trained using a database of existing (human) expert games. The resulting network is a policy network that maps a given board setup (state) to the chess move to be executed in that state that more likely leads to a win (action).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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