Optimisation with genetic algorithm method using ANN for approximation of input parameters

Hi guys,

I am making an optimal design. I have 3 inputs and 1 output. My goal is to maximise the output. There are like one million possibilities and it takes like 5 hours to test each possibility. So since I can`t test all the possibilities, I want to make an optimisation where I use ANN for approximation of the inputs and Genetic algorithm to select the maximum output. So, I will create a response surface (or something like this) using ANN, and select the maximum from the response surface using GA.

1- So it will be like that: I will make like 20 tests with rondom inputs and get outputs for each test. After that, I will approximate my inputs using the 20 tests (its like generating 1000 tests from 20 tests). At the end, Genetic Algorithm will find the maximum output. How can I do it in MATLAB? Can you advise me some tutorial maybe?

2- If I can make accomplish the #1 once, I am thinking to do it like a loop. So, after getting one optimal design, I will put it together with first population, my population number will be 21. I will make again approximation with ANN, find maximum output with GA and test it again. I will make the same loop like 10 times. How can I do it? Actually it should not be difficult after accomplishing #1, I will repeat the same loop with a population that has more rondom test number.

Your help is very crucial for me, can you show me a way how to do it in MATLAB? Any material could be helpful. Please consider that I am relatively new to MATLAB.

Please see the attached graph to understand better what I say (ElFarra et al. 2012).

Thank you very much.

[sara janahi]

Hi

I just saw your question and I'm facing the same problem, so have find a way to do it?if yes please share with me the procedure.

Thanks in advance