# DarrylStein

Member

46

653 Good

• Rank
Member
1. ## MCTS AI development for "Tammany Hall"

Thanks for posting... very insightful
2. ## DarrylStein

Whats worse than losing your wallet? Finding it after reordering all your cards :(
3. ## AI for chess similar games

Think about how many options you have available per move (if the board is almost infiinite - then potentially you have almost an infinite amount of options), can those options be limited in some manner - are only a subset of them reasonable to inspect.  If you can answer those questions it may lead you closer to an answer on how to solve it.  For instance an almost infinite amount of viable options instantly rules out adversarial search like minimax/ab-pruning.
4. ## [Answered] Real numbers (constants) in genetic programming

The implementation that I used for genetic programming was representing the program as a tree.  Each node in the the tree had an INode interface with a method 'Evaluate( INode[] params)' Then I had concrete implementations of nodes for each of the different operators I wanted to represent   operator+ operator* sqrt while(...) if(...) and so on...   I also had a literal operator which just returned a bounded value which was generated randomly at instantiation of the tree (through initial creation or mutation) so at runtime I would create a concrete class through c# IL  e.g.   class literal5 : INode {      INode Execute(INode[] params)     {             return 5.0;     } }   hope that helps
5. ## Need AI Help dealing with trends in Random Numbers

using System; using System.Collections.Generic; using System.Threading; public class ColumnPick { private static readonly Dictionary<int, int> NumberCount = new Dictionary<int, int>(); public enum Face { Zero = 0, One = 1, Two, Three, Four, Five, Six, Seven, Eight, Nine }; public static void Main() { for (int i = 0; i < 10; i++) NumberCount[i] = 0; const int maxIndex = 15000; var rand = new Random(); int oldHighest = 0; for (int i = 0; i < maxIndex; i++) { int next = rand.Next(0, 10); NumberCount[next]++; Console.Clear(); for (int j = 0; j < 10; j++) Console.WriteLine("{0}: {1}", Enum.GetName(typeof(Face), j), NumberCount[j]); if (NumberCount[next] > oldHighest) { oldHighest = NumberCount[next]; Thread.Sleep(500); } } Console.Write("Press any key to continue..."); Console.ReadKey(); } } I've simplified the code. Now to find the top x%, just sort the dictionary and take the top x% x 10 number of buckets.
6. ## Latest trends in procedural quest generation

There are a few articles here that might be of interest. http://www.gamesbyangelina.org/  Maybe the Jul 20 one?
7. ## Negascout behaves oddly

One thing to look for when the AI is not taking the immediate win.  Is to see if the evaluation function discriminates between depth when it finds a win.  For instance, if a win is found at depth 2 and one at depth 5. It might choose the one at depth 5 which looks odd to a human - even if the win is still forced.
8. ## Speeding up evaluation function

Have you done any profiling on the code to see if there are particular areas that are bottlenecking?
9. ## Bomberman AI approach

While there are multiple player versions of minimax (namely MaxN and Paranoid search), it doesn't handle simultaneous moves very well.  Past researchers have introduced an artificial ordering when simultaneous moves are needed, which has its own problems.
10. ## Heuristic for minimax - Board Game 'Odd'

wow - nice problem to solve.  I think minimax certainly will be useful as you start approaching the end of the game, for two main reasons. The branching factor reduces to reasonable numbers for minimax to handle well.  And there appears to be opportunity to use a quiescent search as it appears that many forcing moves will become apparent when there is an opportunity to block merging (or creation in a limited space) of groups (which I assume becomes very important near the end of the game).
11. ## How to weaken AI?

Whilst this works with just straight minimax, it fails once you start implementing even simple search enhancements. i.e. it would be doubtfully useful with alphabeta and most likely useless with PVS
12. ## How to weaken AI?

Another possibility is to reduce the features in your evaluation function e.g. In chess a new player probably wouldn't understand concepts like passed pawns and outposts whereas a more advanced player would
13. ## Transposition table

It appears that most of these are caused by interactions with other search enhancements.  Surely for the purpose of proving that an implementation is correct, you would disable these enhancements initially?
14. ## Transposition table

One thing I've done is to perform each search with TT on and TT off and then check that the same result is being returned.  The TT shouldn't affect the result (only in extremely rare instances) compared to a normal search - it should only affect the speed.
15. ## Negascout's evaluation function

Also consider that some games just aren't symmetrical (eg Fox and Hounds) and the evaluation will reflect that