search

RasmusBrostroem / ConnectFourRL

0 stars 0 forks source link

Enable use of other rewards for `TDAgent` #100

Closed jbirkesteen closed 1 year ago

jbirkesteen commented 1 year ago

incremental_update() was with the commits in #98 hard-coded towards only depending on the win reward of the agent. Change it such that loss or not-finished rewards can also be used in the update rule.

Further ideas for reward systems are mentioned in the original thread and could be implemented/tried out in a separate issue.

Originally posted by @jbirkesteen in https://github.com/RasmusBrostroem/ConnectFourRL/discussions/43#discussioncomment-6712889

jbirkesteen commented 1 year ago

Woops, already wrote this issue.