MATLAB Central Newsreader - Total Matlab newb needs your help

Total Matlab newb needs your help

Mike Rayner — Wed, 19 Nov 2008 11:15:03 -0500

Hi , I have been asked to design a neural network in matlab that takes vectors of handwritten digits as the training data. Then I have been given some testing data to see if the network works. The data is of the form 47,100, 27, 81, 57, 37, 26, 0, 0, 23, 56, 53,100, 90, 40, 98, 8 for the number 8 and I have over 7000 of these. I have absolutely no idea how to get matlab to do this as I have only just started using it any useful guides or links could really help as I am really stressing over it.

Thanks in advance

Re: Total Matlab newb needs your help

David — Wed, 19 Nov 2008 11:28:02 -0500

"Mike Rayner" <mikerayner@one-place.co.uk> wrote in message <gg0sfn$8um$1@fred.mathworks.com>...
> Hi , I have been asked to design a neural network in matlab that takes vectors of handwritten digits as the training data. Then I have been given some testing data to see if the network works. The data is of the form 47,100, 27, 81, 57, 37, 26, 0, 0, 23, 56, 53,100, 90, 40, 98, 8 for the number 8 and I have over 7000 of these. I have absolutely no idea how to get matlab to do this as I have only just started using it any useful guides or links could really help as I am really stressing over it.
>
> Thanks in advance

ok, we get the idea, we don't need it 3 times.

Re: Total Matlab newb needs your help

Adam — Wed, 19 Nov 2008 13:36:02 -0500

"David" <dave@bigcompany.com> wrote in message <gg0t82$gks$1@fred.mathworks.com>...
> "Mike Rayner" <mikerayner@one-place.co.uk> wrote in message <gg0sfn$8um$1@fred.mathworks.com>...
> > Hi , I have been asked to design a neural network in matlab that takes vectors of handwritten digits as the training data. Then I have been given some testing data to see if the network works. The data is of the form 47,100, 27, 81, 57, 37, 26, 0, 0, 23, 56, 53,100, 90, 40, 98, 8 for the number 8 and I have over 7000 of these. I have absolutely no idea how to get matlab to do this as I have only just started using it any useful guides or links could really help as I am really stressing over it.
> >
> > Thanks in advance
>
> ok, we get the idea, we don't need it 3 times.

This isn't helpful.
People new to the interface tend to hit submit more than once (say it doesn't respond immediately) not realizing it will double post. Whatever, it happens.

Mike, the scope of what you're asking is quite large. Do you have any programming experience? Have you done anything with Neural networks.

If this was something that could be explained in a couple lines don't you think it would have been done already?

~Adam

Re: Total Matlab newb needs your help

Mike Rayner — Wed, 19 Nov 2008 17:04:02 -0500

yes I have plenty of programming experience & appreciate this isn't something I am going to be able to work out in a few hours. I have been giving a bit of code which is in c for 'finding zero'.
which goes like :-for i=1:size(s,1)
    if s(i,17)==0
        s(i,17)=1;
    else s(i,17)=0;
    end
end
P=s(:,1:16);
P=P';
T=s(:,17);
T=T';

I need to start heading in the right direction with this task at the moment I am going nowhere slowly.

Thanks in advance for any further help anyone can give.

Re: Total Matlab newb needs your help

Adam — Wed, 19 Nov 2008 17:24:02 -0500

"Mike Rayner" <mikerayner@one-place.co.uk> wrote in message <gg1gu2$8al$1@fred.mathworks.com>...
> yes I have plenty of programming experience & appreciate this isn't something I am going to be able to work out in a few hours. I have been giving a bit of code which is in c for 'finding zero'.
> which goes like :-for i=1:size(s,1)
> if s(i,17)==0
> s(i,17)=1;
> else s(i,17)=0;
> end
> end
> P=s(:,1:16);
> P=P';
> T=s(:,17);
> T=T';

Reading this would help, Vectorizing in Matlab:
http://www.mathworks.com/support/tech-notes/1100/1109.html

Your code could be condenced to, (I believe) ...
P = s(:,16) .';
T = ~s(:,17) .';

I don't really know anything about neural networks, but if you post code people will help you optimize, improve, matlabize it.

~Adam