Prev: what is quantization level of matlab filter implementation
Next: saving the overlayed figure with the alphadata properties
From: cetin tozkoparan on 26 Apr 2010 10:35
The function is y = 1/2 * x^3 - 1/5 * x^2 - 3*x +20
Input value range is [-10; 10]
center vector is [-8; -5; -2; 0; 2; 5; 8]

Plot the output of RBFNN and the function. (Sigma = 0.2)

if anyone can give an starting idea that how to plot this two function, i am really thankful.
From: Greg Heath on 27 Apr 2010 00:37
On Apr 26, 10:35 am, "cetin tozkoparan" <cetintozkopa...(a)yahoo.com>
wrote:
> The function is y = 1/2 * x^3 - 1/5 * x^2 - 3*x +20
> Input value range is [-10; 10]
> center vector is [-8; -5; -2; 0; 2; 5; 8]
>
> Plot the output of RBFNN and the function. (Sigma = 0.2)
>
> if anyone can give an starting idea that how to plot this two function, i am really thankful.  

To design the net see

doc newrbe
help newrbe

and use

p = [-8; -5; -2; 0; 2; 5; 8]' % Notice the transpose
t = 1/2 * p.^3 - 1/5 * p.^2 - 3*p +20

Then, given

x = -10:0.25:10;

calculate the true output

y = 1/2 * x.^3 - 1/5 * x.^2 - 3*x +20

and compare it with the corresponding nn output

z = sim(net,x)

MSE = mse(z-y) % mean-squared error

plot(x,y,'b',x,z,'r'))

Hope this helps.

Greg
 | 
Pages: 1
Prev: what is quantization level of matlab filter implementation
Next: saving the overlayed figure with the alphadata properties