On Mon, Aug 31, 2009 at 11:07 PM, Joseph Heled <[email protected]> wrote:
> I am not sure if you are missing the point or me. Being "more accurate" may > be nice but unless the training is repeated it is probably harmfull. > > The new sigmoid should* give the exact same values as previous implementation. It's just doing four and four values at the time. In that way we only need one float to int conversion (and scaling and adding one and reciprocal) for each 4 value. Table lookups are done one by one though! It's even the exact same table! * /Should/ means that it should give the same answer. I have tested a lot of inputs and got the desired output, however..... there can be surprises. -Øystein
_______________________________________________ Bug-gnubg mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-gnubg
