Author: Vasik Rajlich
Date: 08:25:53 02/06/06
Go up one level in this thread
On February 06, 2006 at 08:35:22, Steve Lim wrote: >On February 05, 2006 at 09:42:16, Torstein Hall wrote: > >>On February 05, 2006 at 06:15:34, Vasik Rajlich wrote: >> >>>On February 04, 2006 at 07:23:06, Michael K. wrote: >>> >>>>Hy Guys! >>>> >>>>Anybody knows, when Beta 14 is coming out? >>>> >>>>Regards! >>>>Michael >>> >>>Rough and completely unofficial schedule is: >>> >>>Feb 8: Beta 13c - straightens out hash table issues >>>Feb 14: Beta 14 - endgame knowledge >>>Feb 16: Beta 15 - persistent hash table (ie. written to disk), plus tool to >>>merge persistent hash tables >> >>Is this some kind of position learning? So if you select the "remember analyzis" >>option in Rybka, the program will still remember it after a restart? >> >>Torstein >> >>> >>>Vas > >I hope this is somewhat similar to the retrograde analysis of shredder. By far >the best implementation I have seen. I do not know about Yace. > >Steve. It's quite simple - deep hash entries will be stored in a hash table on the disk. This can be used also to boost playing strength - once an opening position has been analyzed to a certain depth, it does not need to be analyzed again, provided that the opening book is done and has passed the position to the engine. This has the nice property that two such hash tables can be merged, and the merged data can be used without any adverse affects. (In constrast, true learning always has some potential adverse effects.) Re. "remember analysis" mode, a more intelligent version of this will be the default during infinite (ie. manual) analysis in Rybka 1.2. If the user just jumps back and forth through some variations, the major stuff will be remembered automatically. Vas
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.