> So there's a separate program for developing NNs? I'd be interested > in having a look. What do I need to do?
Yes, Joseph has developed some tools. If your CVSROOT environment variable is set to the usual right value. I think you can just type: cvs co gnubg-nn If you don't use cvs, I can mail you the code. If you can't wait, you can browse it here: http://cvs.savannah.gnu.org/viewcvs/gnubg/gnubg-nn/ The code uses a lot of python. > Am I right in thinking that the 5-node pruning nets do not use SSE > vectorisation. At one point, you were considering implementing this. > Did anything come of it? No, I don't think there is much to gain. > IIRC, you also mentioned increasing the hidden nodes to 8 - because > the loops have to be in multiples of 4. Exactly. With five hidden nodes it's just one multiplication with a sse register and then an additional multiplication in an ordinary register. Increasing to 8 gives a two multiplications in a sse register. I guess it will be about the same computation speed as with 5 hidden nodes, but I guess it won't improve the playing strength much. However, I may improve the code structure. > LINEARITY I'll get back to this. -Øystein
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Bug-gnubg mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-gnubg
