Hello Everyone!
I am trying to optimize two objectives functions (f1, f2) based on two decision variables (x1, x2) with none equality and inequality condition. Only lower and upper bounds limits are given. It seems Multi-Objective Genetic Algorithm (MOGA) Matlab Apps works well and generates Pareto front and gives values of f1, f2, x1 and x2. However, if I manually check the given objective functions values based on particular set of x1 and x2, These f1 and f2 do not match with MOGA's suggested values of f1 and f2. Where may be the problem? All other settings in MOGA have been selected default values. Any help or suggestion shall be highly appreciated. Thanks
Dear Mr. Arya,
I guess you have problem in setting the parameters in MATLAB. I do not advise using MOGA in MATLAB because it is hard or difficult to control parameters setting. I am not sure, but it could be because you did not set constraints of equality and inequality and you were satisfied with upper and lower bounds.
A multi-objective optimization problem can formally be stated as
Find the vector v = [v1, v2....,vn]T
of decision variables
which will satisfy a number of equality and inequality constraints and optimizes the following vector function: f(v) = [f1(v), f2(v),...,fk(v)].
The constraints define the feasible region F that contains all the admissible solutions. Any solution outside this region is
inadmissible since it violates one or more constraints.
see the following literature for more details:
(1) Optimization of spectral signatures selection using multi-objective genetic algorithms. Available from: https://www.researchgate.net/publication/221008616_Optimization_of_spectral_signatures_selection_using_multi-objective_genetic_algorithms
Also you cans see the following important book:
K. Deb, Multi-objective Optimization Using Evolutionary Algorithms, England: John Wiley and Sons, Ltd, 2001.
I think this is due to some random selection factors in MOGA, when you calculate manually you choose one specific factor, but Matlab take random value for some factors.
Dear Prof. Awad,
Thank you very much for sparing time to reply your valuable suggestions. You guess might be correct, since I have not set any set constraints of equality and inequality. However, I feel, MOGA in MATLAB app gives many choices of controlling the parameters such as population type and size, selection criteria, mutation probability, crossover and so on.
I am also thankful for sharing your published article with formal definition of MOGA problem statement. It is really helpful for viewing more insight into my problem.
By the way, if you can find some time to look into my issue, I can share my code of hardly 8-10 lines for setting up the constraints.
With regards and thanks
(Subhash Arya)
Dear Mr. Jabir,
Thank you for drawing attention towards selection factors. I assume initially "tournament' selection which does not show any error while simulating the program, however, if I change selection from "tournament'" to custom "@functionname" then it shows an error that "too many input arguments". In that case how to resolve the issue. Since I have two input parameters, of which I have supplying lower and upper bound. Is there any other things of "selection factor" which I could not trace where you might be addressing/indicating? With regards and Thanks: Subhash Arya
Dear Mr. Arya,
Please send your code so I can look at them.
Best wishes
- Badges
- Science topic