Jason House wrote:
Maybe the best method is to mix the top down
style of MTD(f) to drive localized alpha beta searches.
MTD(f) *is* a sequence of alpha-beta searches.
I definitely don't have all the answers.
MTD(f) doesn't parallelize any better than normal alpha-beta. The only
"advantage" it has is that you theorethically don't need to transmit
beta. But that's quite simply not the problem when parallelizing a search.
Therefore, most modern chess engines still implement Negascout,
which is considered by many chess programmers to be the best search
algorithm in practice.
Modern chess programs also run on single core or very few cores.
This is because most of us don't have access to government
supercomputers. The ones who did (Diep and Zappa) used PVS.
The *paper* about MTD(f) is extremely interesting because it shows that
many best-first algorithms can be rewritten as depth-first algorithms.
It happened for SSS, it happened for proof-number search.
Who will make it happen for UCT?
--
GCP
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
