[quote="Milos"]
Read detailed description of MCTS from AGZ paper and it would be clear to you too.
[/quote]
Milos,
Where can I find "AGZ paper"?
Thanks
Robert
AlphaZero is not like other chess programs
Moderators: hgm, Rebel, chrisw
-
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
-
- Posts: 405
- Joined: Sat Jul 02, 2011 10:49 pm
Re: AlphaZero is not like other chess programs
this may be it ...corres wrote:Milos,Milos wrote:
Read detailed description of MCTS from AGZ paper and it would be clear to you too.
Where can I find "AGZ paper"?
Thanks
Robert
https://deepmind.com/documents/119/agz_ ... nature.pdf
it will ask you if you want to save or open the pdf file ...
-
- Posts: 1301
- Joined: Sun Mar 12, 2006 6:46 pm
- Location: Kelowna
- Full name: Tony Mokonen
Re: AlphaZero is not like other chess programs
speaking of PDFs...
http://incompleteideas.net/book/bookdraft2017nov5.pdf
Reinforcement Learning: An Introduction, 2nd edition complete draft, almost the final version, except no index. Get it while it's free.
http://incompleteideas.net/book/bookdraft2017nov5.pdf
Reinforcement Learning: An Introduction, 2nd edition complete draft, almost the final version, except no index. Get it while it's free.
-
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: AlphaZero is not like other chess programs
[quote="shrapnel"]
You are again nit-picking, Szabo.
[/quote]
MR. Dharan,
Without nit-picking a well-working machine never can be made.
When the machine is working well you as a user can say any high-flown statement about it.
I am afraid you do not know that a neural network is a kinds of memory also.
The process of learning yields much million of chess games what influences the state of the neural network. So if AlphaZero meets a situation what it evaluated during teaching it can find the good winning move. This situation is like as a normal engine use an opening book with very high depth.
Bests.
Robert
You are again nit-picking, Szabo.
[/quote]
MR. Dharan,
Without nit-picking a well-working machine never can be made.
When the machine is working well you as a user can say any high-flown statement about it.
I am afraid you do not know that a neural network is a kinds of memory also.
The process of learning yields much million of chess games what influences the state of the neural network. So if AlphaZero meets a situation what it evaluated during teaching it can find the good winning move. This situation is like as a normal engine use an opening book with very high depth.
Bests.
Robert
-
- Posts: 18764
- Joined: Thu Mar 09, 2006 6:40 pm
- Location: US of Europe, germany
- Full name: Thorsten Czub
Re: AlphaZero is not like other chess programs
To me it looks like stockfish, a very strong chess program so far,
Was beaten because AZ played speculative sacs and moves stockfish could not defend against within its search depth.
It ate each sac because it had no clue that the sac is Sound or will lead into a win for the opponent.
E.g. the game with Re1 .
Stockfish can be seen as a materialistic program eating a sacced piece to get a higher material value.
AZ meanwhile seems to play not for making best moves that lead to material wins, but to create and run a plan.
For the plan it sacs pieces.
So stockfish plays moves in a position that is exact within interval search depth.
But AZ plays plans in a completely different interval then stockfishs interval.
This plan making chess with speculative sacs is IMO very similar to human chess.
Was beaten because AZ played speculative sacs and moves stockfish could not defend against within its search depth.
It ate each sac because it had no clue that the sac is Sound or will lead into a win for the opponent.
E.g. the game with Re1 .
Stockfish can be seen as a materialistic program eating a sacced piece to get a higher material value.
AZ meanwhile seems to play not for making best moves that lead to material wins, but to create and run a plan.
For the plan it sacs pieces.
So stockfish plays moves in a position that is exact within interval search depth.
But AZ plays plans in a completely different interval then stockfishs interval.
This plan making chess with speculative sacs is IMO very similar to human chess.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
Here we have a fairy tale of the day after tomorrow....
-
- Posts: 1339
- Joined: Fri Nov 02, 2012 9:43 am
- Location: New Delhi, India
Re: AlphaZero is not like other chess programs
Immortal Zugzwang game.
https://www.youtube.com/watch?v=lFXJWPhDsSY
https://www.youtube.com/watch?v=lFXJWPhDsSY
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
-
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: AlphaZero is not like other chess programs
[quote="Uri Blass"]
probability is certainly not different from thought of human.
.[/quote]
Probability in AlphaZero is certainly different from thought of human.
During teaching and playing the machine counts the possible move number and counts the occurrence of every each move at the evaluated position. Based these numbers it calculates the possibility of occurrence for every each possible moves and it stores the result. In the played game AlphaZero moves the figure with most higher probability.
probability is certainly not different from thought of human.
.[/quote]
Probability in AlphaZero is certainly different from thought of human.
During teaching and playing the machine counts the possible move number and counts the occurrence of every each move at the evaluated position. Based these numbers it calculates the possibility of occurrence for every each possible moves and it stores the result. In the played game AlphaZero moves the figure with most higher probability.
-
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: AlphaZero is not like other chess programs
[quote="pilgrimdan"]
this may be it ...
https://deepmind.com/documents/119/agz_ ... nature.pdf
it will ask you if you want to save or open the pdf file ...
[/quote]
Thanks for it.
Robert
this may be it ...
https://deepmind.com/documents/119/agz_ ... nature.pdf
it will ask you if you want to save or open the pdf file ...
[/quote]
Thanks for it.
Robert
-
- Posts: 1339
- Joined: Fri Nov 02, 2012 9:43 am
- Location: New Delhi, India
Re: AlphaZero is not like other chess programs
Yes, that's what one of the Commentators said. Stockfish places a very high value on Material, probably because it was programmed that way. So poor Stockfish was probably already laboring under a handicap, while AlphaZero had no such handicaps as it learned all about chess itself within a few hours and was not influenced by the prejudices of humans .mclane wrote:
Stockfish can be seen as a materialistic program eating a sacced piece to get a higher material value.
AZ meanwhile seems to play not for making best moves that lead to material wins, but to create and run a plan.
For the plan it sacs pieces.
This plan making chess with speculative sacs is IMO very similar to human chess.
Alphazero only wanted to win the game and it did not care how many Pieces it sacrificed in the Process.
That is indeed the very Foundation of chess.
At the same time, the sacrifices it made were not like a human player would make, or Stockfish would have seen through them.
The sacrifices Alphazero made were so deep, that even a powerful program like Stockfish could not see through them and was forced to accept them.
That in itself should give pause for thought to the most vociferous defenders of Stockfish.
https://www.youtube.com/watch?v=lb3_eRNoH_w
Last edited by shrapnel on Sun Dec 10, 2017 10:41 am, edited 3 times in total.
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
-
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: AlphaZero is not like other chess programs
[quote="tmokonen"]speaking of PDFs...
http://incompleteideas.net/book/bookdraft2017nov5.pdf
Reinforcement Learning: An Introduction, 2nd edition complete draft, almost the final version, except no index. Get it while it's free.
[/quote]
Thanks.
Robert
http://incompleteideas.net/book/bookdraft2017nov5.pdf
Reinforcement Learning: An Introduction, 2nd edition complete draft, almost the final version, except no index. Get it while it's free.
[/quote]
Thanks.
Robert