I have a cost function need to optimize it using meta-heuristic algorithms. But the function contains both variables and constants how can i solve it?
The problem is how to build the cost function now the cost contains variables and constants so how the output will be function of the constants
how can constants can be defined and the optimizer how can work if the cost contains many undefined constants like C , K < M
I am not sure about your encoding scheme. A graph one could help you. The input can be a constant or a variable. I would suggest you read about Cartesian Genetic Programming.
I think it is a problem how you define your objective. Could you make your problem clearer?
I would like to recommend you the following article uploaded on RG.
https://www.researchgate.net/publication/296636431_Codes_in_MATLAB_for_Particle_Swarm_Optimization?ev=prf_pub
Code Codes in MATLAB for Particle Swarm Optimization
I've heard this question a few times in the past, and in all cases the constants turned out to be not real constants. The mindset was to optimise f(a,b,c,d) for a range of d, and the conclusion was that a scan needed to be made over d. Not so. This d in fact always turns out to be something like price, power, weight or length, and in all cases I have seen this is in fact a separate and conflicting objective. The solution is to realise this, and see the problem NOT as a parameter scan over the 'constant' d, but as a multi-objective simulation where you optimise BOTH f(a,b,c,d) and d. This new problem you can solve effectively with multi-objective optimisation techniques such as NGSA. The additional advantage is that rephrasing as multi-objective actually requires far less function evaluation since this allows the internal algorithms to exploit the fact that there often is a lot of overlap in the optimisation of f(a,b,c,d) for neighbouring d.
dear Bas Van der Geer you answer must be analysed carefully, thank you dr i will focus on it and get back to you again
- Badges
- Science topic