I have an interest to define rules for a heuristic, which is obtained from final solutions of meta heuristics approach. The concept may be to extract general rules used to generate final solutions obtained using the meta heuristics approach.
Can anyone share ideas or papers about this topic? Thank you.
Dear Ivan:
Most of the heuristics constitutes different rules to make evolve in a promising way of one or several solution codes populations . Starting from this concept I developed the Integration of Variables Method, that you can find, together with several applications in the attached work. I am sure it will be useful for the solution of your problem.
Best wishes,
Jose Arzola Ruiz
Hello
I suggest you to consult and view all of this documents
SOLTANI, Roya et SADJADI, Seyed Jafar. Scheduling trucks in cross-docking systems: A robust meta-heuristics approach. Transportation Research Part E: Logistics and Transportation Review, 2010, vol. 46, no 5, p. 650-666.
OSMAN, Ibrahim H. et KELLY, James P. (ed.). Meta-heuristics: theory and applications. Springer Science & Business Media, 2012.
Thank you
Hi Ivan.
Your idea is interesting, but probably you should look outside the metaheuristics literature.
I think what you want to do could be achieved from machine learning algorithms, and if your goal is to have something that slightly resemble rules, you could try decision trees (I'd recommend random forests).
But most likely your analysis should consider the whole search history from your metaheuristic, to have enough observations such that the algorithm could learn to distinguish between promising and non-promising regions.
If you want that all in a single algorithm, use EDAs and check for the outcome model :D
Professor Leonardo,
Thank you for reminding me about decision trees. It may be a good approach to obtain rules for generating good solutions. Next step may be to check whether decision trees could obtain better solution than the ones generated using rules obtained from the meta heuristic approach (if it is possible).
Could you give any example about machine learning methods or papers, which generates some rules?
Could you please tell me what EDA is?
Thank you.
An EDA is an estimation of distribution algorithm. It tries to evolve local models (candidate solutions) to understand how the search space can be modeled as function of the decision variables, so the best solution in an EDA is kinda what you're looking for (except that they are probability distributions, nor trees neither rules).
My background is not in machine learning, so I wouldn't know specific papers to point you to, but start with EDAs and then you can move on to other approaches.
Dear Ivan
i think these article and thesis will help you
https://www.researchgate.net/publication/304674751_FUZZY_CONTROLLER_DESIGN_FOR_DC_MOTOR_USING_ANT_COLONY_OPTIMIZATION_ALGORITHM
https://www.researchgate.net/publication/305701440_Improving_Fuzzy_Controller_Performance_by_Using_Ant_Colony_Optimization_Algorithms?ev=prf_pub
Article FUZZY CONTROLLER DESIGN FOR DC MOTOR USING ANT COLONY OPTIMI...
Thesis Improving Fuzzy Controller Performance by Using Ant Colony O...
Thank you Professor Sameh Abd El-haleem Mohammad.
The articles will surely help.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling