Error of Generated fitness function by Neural Network Fitting Tool for Genetic Algorithm: your fitness function must return a scalar value

Hi Guys. I am dummy about Matlab programming and code writing and I just know about toolboxes. I have a problem with the fitness function generated by Neural Network Fitting toolbox for GA Optimization toolbox in MATLAB R2015. I have developed a neural network through nftool and then generated the ANN as a cost function for Genetic Algorithm for being called through Global Optimization Toolbox. I saved this code as a Mat-file and called it with @ in GA toolbox but the GA tool box gives this error: "Your fitness function must return a scalar value."

Please kindly help me as this is very urgent for my PhD. Please find attached the generated code.

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/29922-why-your-question-is-not-urgent-or-an-emergency

Answer 1

[Walter Roberson]

Your code is written to be vectorized, taking a 7 x Q input and returning a 1 x Q result. However, if you examine http://www.mathworks.com/help/gads/vectorizing-the-fitness-function.html you will see that if you turn on the vectorized option you need to pass in a matrix with an arbitrary number of rows and read the samples out of the columns -- in other words it would pass in a Q x 7 matrix rather than a 7 x Q matrix.

You do not show the code that you invoke the function with, but if you currently have

ga(@myNeuralNetworkFunction, 7, something)

then you should change it to

options = gaoptiset('Vectorized', 'on');
ga(@(X) myNeuralNetworkFunction(X.').', 7, something, options);

[Seyed Saeed Banihashemi Namini]

Thank you very much Walter.

[Wycliff Dembe]

This worked for me too. Thanks a lot