Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: NEURAL NET QUICKIES AND BEYOND
Replies: 3   Last Post: Sep 28, 2017 3:34 AM

 Messages: [ Previous | Next ]
 Greg Heath Posts: 6,387 Registered: 12/7/04
Re: NEURAL NET QUICKIES AND BEYOND
Posted: Sep 28, 2017 3:34 AM

"Greg Heath" wrote:

> I wondered how I would change the quick documentation examples to make the outputs repeatable ( via rng(0) ) and more understandable ( using

0 <= NMSE <= 1 AND 0 <= PCTERR <= 100
...

> % CLASSIFICATION AND PATTERN RECOGNITION
> tic, x = input; t = target; tind = vec2ind( t ); rng( 0 )
> net = patternnet; [ net tr y e ] = train( net, x, t ); view( net )
> yind = vec2ind( y ); PCTERR = 100 * mean ( tind ~= yind )
> NMSE = mse( e ) / mean( var( t', 1 ) ), time = toc
>
> % REMARKS:
> 1. No plots of inputs, targets and outputs were obtained.
> 2. NMSE was included for classification because training tries to minimize NMSE, not PCTERR.

INCORRECT!

PATTERNET TRIES TO MINIMIZE THE CROSSENTROPY BETWEEN OUTPUT AND TARGET !!!

1. SEARCH FOR CROSSENTROPY IN WIKIPEDIA

2. SEARCH FOR THE MATLAB CROSSENTROPY DOCUMENTATION:

HELP CROSSENTROPY

DOC CROSSENTROPY

Hope this helps.

Greg

Date Subject Author
6/23/17 Greg Heath
6/25/17 Greg Heath
7/14/17 Greg Heath
9/28/17 Greg Heath