TalkChess.com

Madeleine Birchfield wrote: ↑

Fri Oct 16, 2020 5:07 pm

And so the only real difference between handcrafted evaluation and neural networks is handcrafted features discovered/implemented by humans vs automated feature discovery/implementation.

this is approximately true...

But it's a difference that makes the difference... a BIG difference...

Is it possible to tune a handcrafted evaluation function whose values all start at zero using unsupervised reinforcement tuning? And if yes, how does this differ from training a neural network from zero using unsupervised reinforcement training? yes... it was done before, and it works... but you us...

Leto wrote: ↑

Sun Oct 11, 2020 4:55 pm

Practically no benefit in that result.

There IS a benefit, it shows something is "fishy"... (pun intended)

Because of all the posts about NNUE, evaluation and related things in the last few weeks, I was thinking about something. Basically, an evaluation term is a fixed or calculated number that is added to the base value (or subtracted if the number is negative). We call such a number BISHOP_PAIR_BONUS,...

SF doesn't have a closed team of developers. Dozens of people have contributed to it, and the NNUE authors are among them. Perhaps we have a different understanding of what that means. Apparently we have, even though what I wrote is not very complex. I guess in your view 99% of Stockfish is not an ...

https://github.com/AndyGrant/Ethereal/tree/nnue Pushed NNUE support to a branch. You can load in a file. Stockfish weights are preferred. If anyone tests this for a rating list, it should be called "Ethereal 12.75 SF-NNUE". That branch actually works, unlike the Etherlito crap which broke due to me...

I don't know if I'm ripping myself away from something I love for no reason. If I'm doing it because I see the pain in the future. If I'm doing it because I feel the pain now but am too daft to see the light at the end of the tunnel. I need someone to convince me that this hobby is still worth purs...

Ethereal has one of the most readable, cleanest codebases out there... I enjoyed reading it... :) plenty of comments in code, too... it is not only one of the strongest engines, it's code is just beautiful... If I were to recommend to someone to clone/steal an engine codebase, it would be Ethereal.....

My analysis and testing have lead me to conclusion that countermove & follow-up history counters help with move ordering at all-nodes, because cut-nodes are already almost optimally ordered... what happens is that at all-nodes this history counters help with finding out new PV lines... because of su...

I am programming in both. Some functions like initialization functions and engine protocol functions are in C. Search functions will be in assembler. MSVC can compile both kind of files and link them. I know how to access non nested C structures from the .asm file. But nowhere shows me how to acces...