Good day, everyone.
I am trying to create an artificial neural network with Matlab and then optimize it with the genetic algorithm. After training the network, I saved it and created the following two script files.
function fitness = fitnessFunction(x, trainedNetwork)
% Check input values
net = trainedNetwork;
if any(x < -1) || any(x > 1)
error('Input values must be between -1 and +1.');
end
try
outputs = net(x'); % use net instead of trainedNetwork
fitness = -sum(outputs); % Negative sign, since the genetic algorithm thinks it is looking for a minimum
catch ME
rethrow(ME);
end
end
%Install trainedNetwork
load('trainedNetwork.mat', 'trainedNetwork'); % Load the file where the trainedNetwork variable is saved
inputSize = 3;
% Set genetic algorithm settings
% Set lower and upper limits (between -1 and +1)
lb = -ones(1, inputSize); % Lower limit
ub = ones(1, inputSize); % Upper limit
options = optimoptions('ga', 'Display', 'iter', 'PopulationSize', 100, 'MaxGenerations', 50);
% Run the genetic algorithm
[x, fval] = ga(@(x) fitnessFunction(x, trainedNetwork), inputSize, [], [], [], [], [], lb, ub, [], options);
% Best solution and fitness value
bestInputs = x;
bestFitness = -fval; % We return it because we have marked it negative
% Calculate outputs corresponding to the best inputs
bestOutputs = trainedNetwork(bestInputs');
disp('Best inputs:');
disp(bestInputs);
disp('The highest outputs corresponding to these inputs:');
disp(bestOutputs);
When I run the second script, I get the following error.
Array indices must be positive integers or logical values.
Error in fitnessFunction (line 8)
outputs = trainedNetwork(x');
I would be very grateful if you can help me.
Elif Çelik
are you trying to use the GA to maximize the output of NN? you are using the output of NN directly so you need to make sure weather element positive or negative so the fitness work correctly also check the dimensions of trainedNetwork to see if it could be used as array with index or not
Dear Elif Çelik ,
The idea here is to employ the Genetic algorithm to optimize ANN parameters to improve performance. ANN provides the search space and utilizes GA to find the best solution by tuning the weights and biases required to achieve lower error rates. The error between the model output and the exact training data can reach a minimum value by iterating the GA until the desired error is met.
https://fr.mathworks.com/matlabcentral/fileexchange/124600-hybrid-artificial-neural-network-with-genetic-algorithm
Regards,
Shafagat
- Badges
- Science topic
- Similar topics
- Psychology