Help in Neural network Coding?

Good Afternoon I'm new to neural network and MATLAB. I'm try to make neural network to predict Core Facies (e.g. I have 8 facies) in oil field from wireline log (10 wireline logs).

I have many questions and i would be appreciate if someone can help me.

First, How to structure or format my input and output data to be loaded into MatLab? Which type of neural network should I use? Before runing the ANN, how can I rank my wireline log to selected as my input? Finally, the number of target/output, will be 8 outputs of it will be only one?

I would be appreciate any answer.

Regards, Ghalia

Answer 1

[Greg Heath]

 [ I N ] = size(input)    % I = 10
 [ O N ] = size(target)   % O = 8

% Find h by trial and error to minimize MSE

 net            = fitnet(h);     
 [ net tr y e ] = train(net,input,target); % e=target-y
            MSE = mse(e)

For details search

 greg fitnet Ntrials

Hope this helps

Thank you for formally accepting my answer

Greg

Answer 2

[Greg Heath]

You are correct: Use patternnet for classification/pattern-recognition.

Sorry for the oversight. Number of classes c = 8

 [ I N ] = size(input)    % I = 10
 [ c N ] = size(target)   % c = 8

Columns of target are {0,1} unit vectors with

 target      = ind2vec(trueclassindices)
 sum(target) = ones(1,N)

where

 trueclassindices = vec2ind(target)

% Find H by trial and error to minimize MSE (error rates are not differentiable)

 net            = patternnet(h)     
 [ net tr y e ] = train(net,input,target); % e=target-y
            MSE = mse(e)
 predictedclassindices = vec2ind(y)

 totalerrors = predictedclassindices~= trueclassindices; %{0,1}vector

% From this (0,1) vector individual class errors can be obtained. Additional info can be obtained from the training record tr.

For details search

 greg patternnet Ntrials

Hope this helps

Greg

[Oman Wisni]

Hi, I have input = 220x25 and target = 220x1

I'm trying follow your this code

 [ I N ] = size(input)    
 [ O N ] = size(target)   
 

But the resul I got is N = 1, N = 220, O = 220. this is right ?

Answer 3

[Ghalia Al-Alawi]

thanks for your reply. But which neural net should I use for the prediction since I want to predict these facies in uncored wells. Should I use fitnet or pattannet?

Regards, Ghalia