[computer-go] Re: Amsterdam 2007 paper

[Rémi Coulom]

Yamato wrote:
> Thanks for the nice paper.
>
>   
>> I will run cleaner tests for the final version of the paper. I will also 
>> try to test what each feature brings.
>>     
>
> I hope that you do 70k simulations like MoGo, since "90s/game 1CPU"
> depends on the simulation speed and the time distribution.
>
> And I have a question. In section 4.1, you wrote:
>
>   
>> Contiguity to the previous move is a very strong feature (gamma = 23)
>>     
>
> Where has the number 23 come from?

It was learnt by the algorithm. I did not re-use the pattern weights 
presented on Table 1. I ran the learning algorithm again, using a set of 
light-weight features. Contiguity to the previous move got a gamma of 23.

Rémi