Comments on: Stukafest 2009 http://martijn.van.steenbergen.nl/journal/2009/02/19/stukafest-2009/ Just another WordPress weblog Sun, 02 Oct 2011 13:50:03 +0000 http://wordpress.org/?v=2.8.5 hourly 1 By: Joachim Schipper http://martijn.van.steenbergen.nl/journal/2009/02/19/stukafest-2009/comment-page-1/#comment-278 Joachim Schipper Tue, 24 Feb 2009 16:07:28 +0000 http://martijn.van.steenbergen.nl/journal/?p=146#comment-278 Hello Sunny, Are you aware that Martijn's Quoridor implementation can be downloaded via 'Projects', above? (Martijn: it doesn't do anything but show a white canvas with Java-1.6.0_03-p4 on OpenBSD-amd64, although it doesn't pretend to work in that case.) I would recommend reconsidering your choice of hill-climbing as the main algorithm, though. The utility of placing a wall - which is the main computation required of a Quoridor AI - is very irregular as a function of its location. Martijn can tell you the details if he wants - they are not mine to tell - but I can tell you that his implementation doesn't use hill-climbing. Also, I have personally been unable to devise a faster algorithm than A* for the numerous pathfinding steps required (since placing a wall "changes little," you could try to re-use results; but the extra bookkeeping more than cancels the reduced computation.) Finally, this really isn't an appropriate place for your question. Martijn, if you delete it, can you please delete this one as well? Yours, Joachim Hello Sunny,

Are you aware that Martijn’s Quoridor implementation can be downloaded via ‘Projects’, above? (Martijn: it doesn’t do anything but show a white canvas with Java-1.6.0_03-p4 on OpenBSD-amd64, although it doesn’t pretend to work in that case.)

I would recommend reconsidering your choice of hill-climbing as the main algorithm, though. The utility of placing a wall – which is the main computation required of a Quoridor AI – is very irregular as a function of its location.

Martijn can tell you the details if he wants – they are not mine to tell – but I can tell you that his implementation doesn’t use hill-climbing.

Also, I have personally been unable to devise a faster algorithm than A* for the numerous pathfinding steps required (since placing a wall “changes little,” you could try to re-use results; but the extra bookkeeping more than cancels the reduced computation.)

Finally, this really isn’t an appropriate place for your question. Martijn, if you delete it, can you please delete this one as well?

Yours,

Joachim

]]> By: Sunny Chan http://martijn.van.steenbergen.nl/journal/2009/02/19/stukafest-2009/comment-page-1/#comment-271 Sunny Chan Sat, 21 Feb 2009 08:45:02 +0000 http://martijn.van.steenbergen.nl/journal/?p=146#comment-271 Hello. I am a university student in Hong Kong. And I'm very interested in your Quoridor project since I am also working on artificial intelligence in Quoridor. So, would you mind answering me a few simple questions? First, is it possible for you to tell me the algorithm you used in GoalSeeking? For sharing, my project do GoalSeeking by a theory called "Hill-climbing", but it really takes too much time in every goal seeking. Second, in SmartBrain1, would you mind telling me that if the response is mainly produced by Minimax decision making algorithm or is there any algorithm involved? I observed that the decision making in SmartBrain1 is very fast, which seems not to be using Minimax or negamax. Hello. I am a university student in Hong Kong. And I’m very interested in your Quoridor project since I am also working on artificial intelligence in Quoridor. So, would you mind answering me a few simple questions?
First, is it possible for you to tell me the algorithm you used in GoalSeeking? For sharing, my project do GoalSeeking by a theory called “Hill-climbing”, but it really takes too much time in every goal seeking.
Second, in SmartBrain1, would you mind telling me that if the response is mainly produced by Minimax decision making algorithm or is there any algorithm involved? I observed that the decision making in SmartBrain1 is very fast, which seems not to be using Minimax or negamax.

]]>