Hannibal 1.1- Learning Lemming 0.44T - x64

[Sylwy]

Hi all !

1.A co-author of Hannibal 1.1 engine is Mr. Sam Hamilton.
2.Mr.Sam Hamilton is the author of the private chess engine Learning Lemming 0.44T - x64.
3.Learning Lemming chess engine has a very interesting learning function.
4.Does anybody know if this learning function ( or some ideas ) was implemented also in Hannibal 1.1 ?

Thank you a lot in advance !
SilvianR

[Dr.Wael Deeb]

I am also interested in this learning feature and would be very happy to test it in my private rating list....

Actually,I test chess engines with learning switched on and in most of the cases this option pays back pretty good....
Dr.D

[Sylwy]

I am also interested in this learning feature and would be very happy to test it in my private rating list....

Actually,I test chess engines with learning switched on and in most of the cases this option pays back pretty good....
Dr.D

Yes Doc !

After around 1500 games The Baron performs better in my tournaments !
I suspect also Rybka 4 to have a sort of learning function despite the fact that in his installation folder I can't see any new file. But playing againt a new chess engine in the second part of every match Rybka performs much better !

Have a nice weekend my friend !
Regards,
SilvianR

[Dr.Wael Deeb]

I am also interested in this learning feature and would be very happy to test it in my private rating list....

Actually,I test chess engines with learning switched on and in most of the cases this option pays back pretty good....
Dr.D

Yes Doc !

After around 1500 games The Baron performs better in my tournaments !
I suspect also Rybka 4 to have a sort of learning function despite the fact that in his installation folder I can't see any new file. But playing againt a new chess engine in the second part of every match Rybka performs much better !

Have a nice weekend my friend !
Regards,
SilvianR

Thanks Silvian and have a nice weekend yourself
Dr.D

[BubbaTough]

Hi all !

1.A co-author of Hannibal 1.1 engine is Mr. Sam Hamilton.
2.Mr.Sam Hamilton is the author of the private chess engine Learning Lemming 0.44T - x64.
3.Learning Lemming chess engine has a very interesting learning function.
4.Does anybody know if this learning function ( or some ideas ) was implemented also in Hannibal 1.1 ?

Thank you a lot in advance !
SilvianR

Hannibal has no learning. The learning feature in LearningLemming does not happen during the game. Instead, parts of the game to learn from are saved to a file, and then a separate learning program does a post-mortem analysis and saves off conclusions for LearningLemming to use in future games (you can also add a list of games from an upcoming future opponent into the "tolearn" file, which I did on occasion much to the annoyance of some of my ICC computer opponents). Thus, its kind of a pain for people who run tournaments and keep rating lists to use. For example, in OpenWar, Olivier would occasionally just send me the "tolearn" file generated during games, and I would send him back a new "learned" file generated on my computer, which meant no learning would actually happen during his tournaments since this exchange was infrequent.

The other type of learning I played with in LearningLemming is really just parameter tuning, and a rough form of this is now used now in the Hannibal development process. A number of engine developers use this (probably most of the top engines I would guess).

Adding learning to Hannibal is on my todo list, but it has not been very high since I kind of assumed no one would be that interested in using it with Hannibal being so much weaker than the top engines. Only a few book makers (Erdo was always asking for LearningLemming versions, and our book maker Audy wants it) that used to play LearningLemming on ICC to help tune their books have expressed strong interest. For the most part, I have been assuming that the elo I can add in the time it would take me to add learning support is preferable to most people, at least up until the point where Hannibal's positional analysis is at least close to the quality of the very top engines.

-Sam

[Sylwy]

Adding learning to Hannibal is on my todo list, but it has not been very high since I kind of assumed no one would be that interested in using it with Hannibal being so much weaker than the top engines.

-Sam

Hi Sam !

Very interesting the idea behind the learning function of your Learning Lemming ! Nice to see a learning function on Hannibal engine. I'm extremely interested in original ideas than a speedup chess engine.

Regards,
SilvianR