>   This is very interesting! I'd be intrigued to know how well this works
> for you in practice, e.g. how many games are required for basic tuning
> of few [0,1](R values.
>
>
I would say that this number of games scales as d/epsilon^2
with d the dimension and epsilon the required precision; this is I think
reached by the QLR
software (papers on QLR on Rémi Coulom's web page), if there's no problem
with bad conditionning or too complicated functions.
At least there are upper complexity bounds like that :-)
The parallel complexity is very good - which suggests (unsurprisingly) that
using clusters or grids is good for this.
Rémi, your feeling on this ?
Olivier
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to