Prev: Memoryless Bandpass/Baseband NonLinearities
Next: MANET Mobile Ad hoc Network, Source code, AODV
From: George BC on 21 Mar 2010 09:38
Hello!

Does anyone knows how can I draw a curve that separetes the blues points from the red?

the plot is here:

[URL=http://img203.imageshack.us/i/plotb.jpg/][IMG]http://img203.imageshack.us/img203/3379/plotb.jpg[/IMG][/URL]

I used multi-layer perceptron with the backpropagation algorithm (not using any matlab function) and already got the weigths with 99% of success.
From: Matt J on 21 Mar 2010 11:09
"George BC" <georgecase(a)gmail.com> wrote in message <ho57fq$8hq$1(a)fred.mathworks.com>...
> Hello!
>
> Does anyone knows how can I draw a curve that separetes the blues points from the red?
======

For this particular data, it looks like the convex hull boundary of the blue points would form such a curve. Use convhulln
From: George BC on 21 Mar 2010 11:47
Well.. I not quite sure thats the case, isn't any way that I can create the convex hull with the network weights (of the hidden and output layers)?

something like the line separating a set of points used in a single layer perceptron...

tks.
 | 
Pages: 1
Prev: Memoryless Bandpass/Baseband NonLinearities
Next: MANET Mobile Ad hoc Network, Source code, AODV