2
$\begingroup$

I am designing a algoritm for a game using a Monte Carlo tree search AI that I implemented. I play against another player and want to get the best move. Every time I do a move I completely build a new search tree and do as much iterations possible given a certain deadline. Is there a way to use the previous tree that I build or do I always have to build a tree every time I make a move.

our turn: calculate tree beginning from scratch using uct

otherplayer turn: his turn

our turn: calculate tree beginning from scratch using uct

Is there a way I can optimize this So i could do even more iterations/ be faster ? like for exemple:

our turn first: calculate tree beginning from scratch using uct

otherplayer turn: his turn

our turn: use tree we already calculated in our first turn and do more iterations given the values

Or are there other optimalisations I could do?

EDIT: https://stackoverflow.com/questions/47389700/why-does-monte-carlo-tree-search-reset-tree

$\endgroup$
  • $\begingroup$ (Zeroth improvement: use improvement where something gets better, but not optimal.) $\endgroup$ – greybeard Nov 14 at 17:02
  • $\begingroup$ @greybeard I don't realy get what you are trying to say, wouldn't I be able to do first move: calculate tree from scratch next move opponents: second move: follow the path of first move and opponents move and begin not from scratch but from that node ? $\endgroup$ – swaffelay Nov 15 at 8:59
  • $\begingroup$ Merriam-Webster on optimalisation (really). Not forgetting something that took effort to know and may well be valuable sounds prudent. You may want to capitalise UCT, mention what it means here. Generally follow How do I ask a Good Question? $\endgroup$ – greybeard Nov 15 at 16:45
  • $\begingroup$ m8 your comment isn't useful at all $\endgroup$ – swaffelay Nov 15 at 18:39

Your Answer

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

Browse other questions tagged or ask your own question.