I have stuck with using narnet .

I have used the code provided by Greg in newsgroup. My sample code is here:

Input

clc,clear;
plt=0;
X=load('BOD test.txt');
L=length(X),
net=narnet(1:.1,10);
view(net),
[Xs,Xsi,Asi,Ts] = preparets( net, {}, {}, X );
ts = cell2mat(Ts);
plt = plt+1; figure(plt), hold on
plot( 1:L, ts, 'LineWidth', 2 )
rng( 'default' )
[net tr Ys Es Af Xf] = train( net, Xs, Ts, Xsi, Asi );
view( net )
NMSEs = mse( Es ) /var( ts,1 ) 
ys = cell2mat( Ys );
plot( 1:L, ys, 'ro', 'LineWidth', 2 )
axis( [ 0 22 0 1.3 ] )
legend( 'TARGET', 'OUTPUT' )
title( 'OPENLOOP NARNET RESULTS')

And the output is:

L =
    18
NMSEs =
    1.0000

And what's wrong with curve plot:

[Greg Heath]

PLEASE DO NOT USE THE NEWGROUP AND ANSWERS FOR THE SAME PROBLEM!!!

[Image Analyst]

With all due respect professor, I'm not against looking for solutions in multiple places. In the real world you need to solve a problem and you cast a wide net to get all possible leads. You might get one lead in one place and a totally different approach in a different place. My manager would never say to me "Solve this problem but only look in one place." He just wants the job done and doesn't care how I get the job done (as long as it's ethical). On the contrary they want us to look in as many places as practical for a solution, which may reveal a better solution or get a workable solution faster. Granted, you're the only neural network expert here so in this particular case he probably won't find any different approaches, but in general I don't think it's a bad idea because different people participate in different forums and that could lead to different approaches.

[Yeasir Mohammad Akib]

Sir, sorry for my approach. I won't do this again but I was desperate for seeking the answer for my problem. Besides no one is here to solve my problem even if it's my teacher. So I have to seek help from this type of active group. And I think Image Analyst's perspective is right at all.

[Greg Heath]

I understand your point. However, often the descriptions of the same problem from the same poster in the two forums are different.

Consequently I have found myself dizzily ping ponging back and forth between the two.

It is very, very annoying.

Therefore, I will continue, at least for neural nets, to rant my dissatisfaction. However, I will add a neural net qualifier.

Grumpy Greg

Answer 1

[Greg Heath]

net = narnet(1:.1,10);

 1:0.1 is an error

However, I do not think that you have enough data to do what you wish.

Any further communication should be made via the NEWSGROUP.

Greg