Search found 3387 matches

by Sven
Thu Jan 21, 2021 6:22 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Parameter Tuning Algorithm
Replies: 6
Views: 218

Re: Parameter Tuning Algorithm

4a. [...] The basic condition now is that "eds" < "efs". Is that the condition for one single position to survive in the test set? Why eds < efs and not eds <= efs? 4c. [...] deepscore = Epd::getEvaluation(entry); // retrieves the score in the string ... double Tuner::computeError(int deepscore, in...
by Sven
Thu Jan 21, 2021 4:48 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Weird engine behaviour
Replies: 3
Views: 127

Re: Weird engine behaviour

Hi Elias, the version you showed me via PM correctly finds the mate in 3 (= mate in 5 plies!) at depth 5 and does not forget it. Prior to depth 5 it displays Rxe1 as best move. That version has nullmove disabled. So I suspect that you are now testing a version where nullmove is active and does not w...
by Sven
Thu Jan 21, 2021 1:51 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

Your f() function returns values between 0.0 and (RAND_MAX / 2147483648.0), where the latter is about 1.526e-05 for a typical RAND_MAX=32767. Well, apparently the rand() function used by my compiler is not very typical. Of course the very first thing I did is make the program do 100 rand() calls an...
by Sven
Thu Jan 21, 2021 1:38 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

No, it can also change multiple terms at the same time. If you don't do that, you might get stuck in places that are not even local optima. I used the smallest possible step size (as the piece values are integers in centiPawn). To be sure to find a local minimum you would always have to reduce your...
by Sven
Thu Jan 21, 2021 8:27 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

There seems to be some disagreement about what 'Texel tuning' means. As far as I am concerned it is minimizing the MSE in the prediction of the result of the game from which the test positions were taken. That is, finding the global minimum. If you allow yourself to get stuck in a local minimum, yo...
by Sven
Thu Jan 21, 2021 8:24 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

@Michael: The automated Texel tuning does not know about any possible other local minima when it stops at one. "Modifying the parameter set" is what the algorithm does all the time: it increments/decrements eval parameters by the step size and sees whether that results in a lower MSE. It stops if an...
by Sven
Thu Jan 21, 2021 12:22 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

It is unlikely to find "the minimum MSE" with Texel tuning. The algorithm stops at a local minimum of the MSE, and there may be several of that kind. It seems you ignore this fact. A good optimizer should not get stuck in local optima. And note that for terms in which the evaluation is linear (like...
by Sven
Wed Jan 20, 2021 11:52 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

Pio, in your last post it seems you are confusing myself with Michael (Desperado). I did not tune just piece values (except for non-productive testing purposes), I did not write about any special effects of MG values only, and my tuning process did not return parameters close to zero. Also I still t...
by Sven
Wed Jan 20, 2021 6:18 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

"if", "would", "might", did I miss some?

You have a theory about it. Go ahead and implement it, then test it and make it work under real conditions. Then come back. Until then, please accept my apologies that I ignore what you say about this topic.
by Sven
Wed Jan 20, 2021 5:45 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Tapered Evaluation and MSE (Texel Tuning)
Replies: 220
Views: 6041

Re: Tapered Evaluation and MSE (Texel Tuning)

It is unlikely to find "the minimum MSE" with Texel tuning. The algorithm stops at a local minimum of the MSE, and there may be several of that kind. It seems you ignore this fact. A good optimizer should not get stuck in local optima. And note that for terms in which the evaluation is linear (like...