i want to optimize a neural network parameters by genetic algorithm and i get below code from matlab support, but the result is very poor even for a very simple function. how can i improve the result and is there any better code for my porpuse?

function mse_calc = mse_test(x, net, inputs, targets)

% 'x' contains the weights and biases vector

% in row vector form as passed to it by the

% genetic algorithm. This must be transposed

% when being set as the weights and biases

% vector for the network.

% To set the weights and biases vector to the

% one given as input

net = setwb(net, x');

% To evaluate the ouputs based on the given

% weights and biases vector

y = net(inputs);

% Calculating the mean squared error

mse_calc = sum(abs(y-targets)/length(y);

end

% INITIALIZE THE NEURAL NETWORK PROBLEM %

% inputs for the neural net

inputs = (-2:.01:2);

% targets for the neural net

targets = ((inputs).^2)-3;

% number of neurons

n = 4;

% create a neural network

net = feedforwardnet(n);

% configure the neural network for this dataset

net = configure(net, inputs, targets);

% create handle to the MSE_TEST function, that

% calculates MSE

h = @(x) mse_test(x, net, inputs, targets);

% Setting the Genetic Algorithms tolerance for

% minimum change in fitness function before

% terminating algorithm to 1e-8 and displaying

% each iteration's results.

ga_opts = gaoptimset('PopInitRange', [-1;1], 'TolFun', 1e-10,'display','iter');

ga_opts = gaoptimset(ga_opts, 'StallGenLimit', 100, 'FitnessLimit', 1e-5, 'Generations', 400);

% PLEASE NOTE: For a feed-forward network

% with n neurons, 3n+1 quantities are required

% in the weights and biases column vector.

% a. n for the input weights

% b. n for the input biases

% c. n for the output weights

% d. 1 for the output bias

% running the genetic algorithm with desired options

[x, err_ga] = ga(h, 3*n+1, ga_opts);

net = setwb(net, x');

Answer 1

[Greg Heath]

http://www.mathworks.com/matlabcentral/newsreader/view_thread/344888#944621

If you can improve the code please let me know.

Greg

[Saeed]

thanks for the code Greg but unfortunately it did not give satisfactory results as the above code I put. the approach of the two code was similar and the results are the same.

[Walter Roberson]

You might have a function which cannot really be improved. Or you might have a function for which it is difficult to find an improvement.

[Greg Heath]

Walter,

I don't think so.

I think there is something wrong with the MATLAB code.

I have never seen it successfully demonstrated.

Greg

[Darshana Abeyrathna]

Saeed,

Please share if you have found any solution. Because I have the same problem :(

http://www.mathworks.com/matlabcentral/answers/306379-why-it-gives-worse-results-when-i-use-genetic-algorithm-for-training-nns-than-when-i-use-back-propag

Answer 2

[Don Mathis]

In the code you posted, mse_test is not calculating mean squared error. Here is a corrected version that seems to work fine. Paste the whole thing into the MATLAB editor and run it:

% INITIALIZE THE NEURAL NETWORK PROBLEM %

% inputs for the neural net
inputs = (-2:.01:2);

% targets for the neural net
targets = ((inputs).^2)-3;

% number of neurons
n = 4;

% create a neural network
net = feedforwardnet(n);

% configure the neural network for this dataset
net = configure(net, inputs, targets);

% create handle to the MSE_TEST function, that
% calculates MSE
h = @(x) mse_test(x, net, inputs, targets);

% Setting the Genetic Algorithms tolerance for
% minimum change in fitness function before
% terminating algorithm to 1e-8 and displaying
% each iteration's results.

ga_opts = gaoptimset('PopInitRange', [-1;1], 'TolFun', 1e-10,'display','iter');
ga_opts = gaoptimset(ga_opts, 'StallGenLimit', 100, 'FitnessLimit', 1e-5, 'Generations', 400);

% PLEASE NOTE: For a feed-forward network
% with n neurons, 3n+1 quantities are required
% in the weights and biases column vector.
% a. n for the input weights
% b. n for the input biases
% c. n for the output weights
% d. 1 for the output bias

% running the genetic algorithm with desired options
[x, err_ga] = ga(h, 3*n+1, ga_opts);
net = setwb(net, x');

function mse_calc = mse_test(x, net, inputs, targets)
% 'x' contains the weights and biases vector in row vector form as passed
% to it by the genetic algorithm. This must be transposed when being set as
% the weights and biases vector for the network. To set the weights and
% biases vector to the one given as input

net = setwb(net, x');

% To evaluate the ouputs based on the given weights and biases vector
y = net(inputs);

% Calculating the mean squared error
mse_calc = sum((y-targets).^2)/length(y);

end

[Ammy]

I have a similar problem , I want to add an objective function in the above code that I want to optimize with Particle swarm optimization problem, Is there any way ?