I came across these systems. Are they all attempting to solve the algorithm selection problem?
Dear Patricia Ryser-Welch,
I think that most of the algorithm configuration systems are motivated by the following aspects:
1- The algorithm selection problem as stated by John Rice and pointed out by you.
2- The no free lunch theorem which stated that even for best selected algorithm there might be room for improvement.
3- As a side results of the 1 and 2 points, to reduce the development effort and cost needed for developing a new algorithm.
That is very useful and confirm my understanding. Now are these system generating new algorithms? Are they using existing ones and mapping them to instances?
My understanding is that it is the latter. Often they also set the parameters using some statistical or other methods.
You are right in thinking that algorithm configuration systems develop existing ones and mapping them to instances. It is due to the fact that an autonomous configuration framework requires a correct understanding of both the algorithm behavior and the problem at hand.
Recently, we propose a new framework for evolutionary design of meta-heuristic algorithms and its codes are available at the following link. I hope it will be helpful.
Data MGEP Code
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Algorithms