Let's say that a metaheuristic algorithm has three parameters: A, B and C. Let Y denote its performance metric. For example, Y is the average difference between the objective function value of the global optimum and the best solution that the algorithm finds, measured on a number of benchmarks.
If we construct a regression model, e.g. Y = c0 + c1*(A2) +c2*A + c3*B + c4*C, can we determine the values for A, B and C from the model? I know that we can get certain values for the parameters this way, but is this approach actually used in the literature and practice?
It seems fine to me, but I would like to hear other opinions. :)
You could solve the optimization task for different values ofthe parametrs A, B, C and, for the obtained results, elaborate the regression equation with regard to thesel parameters.
I have carried out this procedure in several occasions with the purpose of elaborate "behavior rules" for the operation of technological processes
In addition, model could have any structure, not neccessarilly Y = c0 + c1*(A2) +c2*A + c3*B + c4*C. If you need more information just let me know
I know, the model in the question is just an example. Thank You for Your comments. :)
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Journalism