How to maximise a multiobjective function with Genetic Algorithm as in its original form, It is used to minimise a multibojective function? Is is the correct method to add a negative sign to the two function to maximise multiobejctive (two functions) function?
Yes, you could change the sing of both objetives. Maximize is equivalent to minimize its negative
Thanks Angel Udias for your answer. I am adding negative sign to the two objective functions. But, then I am getting the negative objective function values at the maximum point. So the maximum values for two objectives will be obtained just by removing the negative sign?
Please would you let us know exactly what the functions are that you are trying to maximise to help you better.
You might have to do something like minimise 1/(func1+func2+some constant term)
or
minimise 1/(func1*func1+func2*func2+some constant term)
or minimise 1/(func1*func2+some constant term)
constant terms are needed to stop the above one over functions from becoming infinity
- I have two functions, Func1 and Func2. I want to maximise both functions simultaneously using multiobjective genetic algorithm tool.
Ravita
We need to know the actual equations for Func1 and Func2, stating which are the constant coefficients and which are the variables that can vary.
LS
Loopinder Sood
Can I inbox you the code? Because the functions are lengthy and just to write the two functions will not give you clear understanding about the optimisation problem?
Yes, you could change the sign of both functions and then use any multi-objective optimization algorithm like NSGA-II to obtain the pareto front.
For a nontrivial multi-objective optimization problem, the objective functions are said to be conflicting, and there exists a (possibly infinite) number of Pareto optimal solutions.
I would not try the suggestion above about minimizing a composite function (like 1/(func1*func1+func2*func2+some constant term as suggested above). You will end up with a point that may not be on pareto front.
Thanks Behrouz Ahmadi-Nedushan sir for your help. But I have one more doubt. I am adding negative sign to the two objective functions which are cooling load and COP of the thermoelectric cooler. But, then I am getting the negative objective function values at the optimum point. So the maximum values for two objectives will be obtained just by removing the negative sign?
Yes, Ravita, indeed. As you pointed out, you re minimizing the negative of your objective functions. The maximum you are looking for are then the negatve of your solutions.
Once the solution gets converted, you should calculate the original objective function values (both Func1 and Func2) at the optimized set of control variables. These values of objective functions will be the actually optimized values. However, combination of the two objective functions is the fitness function that may be positive, negative or zero.