From: leila on
how can'i use bp training algorithm with dynamic threshod in the output layer:
this is my function : can you help me :

F(yi)= 1,Yi= Max{y1,y2,?????.ym}
0,Others
From: Greg Heath on
On Apr 24, 7:17 am, leila <leeeila12...(a)live.fr> wrote:
> how can'i use bp training algorithm with dynamic threshod in the output layer:
> this is my function : can you help me :
>
> F(yi)= 1,Yi= Max{y1,y2,?????.ym}
> 0,Others

You can't. You need differentiable activation functions for
backprop and other training algorithms.

See my previous post.

Greg