Learning feature

[Hood]

[="Uly"]
Then you have a twisted definition of learning. In your mind the Stockfish Gran series have learning, in mine, Learning has yet to be implemented in Stockfish.



May be twisted, may be not . It is short time learning. Do you remember all from a base school ?

[Uly]

Learning on 1000 positions would mean saving 1000 Hash tables (maybe 1 TeraByte... )
Not very practical!...

You're confusing the features, Save/Load Hash is NOT learning, because it's also short-term and will be forgotten as soon as Rybka overwrites a hash entry. Rybka 4 still has the Save/Load Hash feature, but the reports I've seen make me think that it's not worth the hassle.

Persistent Hash is a different feature, in where you set it once and Rybka 3 will automatically store all positions analyzed on a single file of the specified size. I have mine set to 256MB, and it's a single file, all positions are stored in it so I don't need to create a new one for each position.

(Is still full of bugs and only recommended for users that know how to workaround them)

@Hood:

I don't remember everything from base school but I remember things from yesterday and really old anecdotes. The preserve analysis feature you talk about is akin to getting amnesia every time you wake up, that's how engines without learning work, their whole memory is erased every time you unload them.

[ernest]

Persistent Hash is a different feature, in where you set it once and Rybka 3 will automatically store all positions analyzed on a single file of the specified size. I have mine set to 256MB, and it's a single file, all positions are stored in it so I don't need to create a new one for each position.

OK thanks, Uly!

[Hood]

I don't remember everything from base school but I remember things from yesterday and really old anecdotes. The preserve analysis feature you talk about is akin to getting amnesia every time you wake up, that's how engines without learning work, their whole memory is erased every time you unload them.

If i made save hash it is not neccessarily the amnesia.

[tano-urayoan]

If i made save hash it is not neccessarily the amnesia.

Have you tried Rybka's save hash before? How did it worked for you? For example you analyzed a position to depth 25 after you unloading the engine and later return to that position did it started analyzing at depth 26?