1
$\begingroup$

Production systems have been used to solve puzzles such as the Tower of Hanoi for years with hard-coded production rules. However, has there been any research in using reinforcement learning to combine basic production rules into more complex ones with the goal of solving a puzzle?

Note that I've asked a similar question on Cogsci.SE, but with a greater cognitive perspective. In my opinion, the two perspectives are different enough to merit semi-cross-posting.

$\endgroup$
1
$\begingroup$

this writeup by Kenan Deen solves the Tower of Hanoi problem with reinforcement learning and Markoff decision processes.

there is some other research on using genetic programming in "solving" (successfully playing) video games. these approaches do find "subroutines" that can be regarded as subgoals. for example, in playing pacman, an algorithm might have a subroutine/ subgoal that eats ghosts if they are blue, otherwise, run away from them. the sample algorithms are sometimes encoded as LISP-like trees. see eg

$\endgroup$

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.