[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [computer-go] Designing faster, better influence functions



Gunnar Farnebäck wrote:
My guess is that you can't perform any calculations relevant for an
influence function in the Fourier domain and even if you could it
wouldn't improve performance. But please elaborate if you think I'm
wrong.
Well, the point is that convolutions in the spatial domain are multiplications in the Fourier domain. Multiplications are of course faster. The question is how much time extra the FFT /IFFT transforms take. I'm pretty sure others have already tested this, so I was hoping somebody would remember the timing details.

BTW with knowledge of life and death available, simple linear filtering works quite well.

Erik
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/