This question is framed just to get an idea about the use of meta-heuristics. How to select the meta-heuristics for any problem, or use of which meta-heuristics will give the better solution. If responses are from supply chain, scheduling and vehicle routing areas would be quite interesting and of much use. Your suggestion are important and welcome.
Meta-heuristics are not problem-specific !
every one claims that "global optimization algorithm". . But, finally the algorithm which " Minimize/Maximize ur Objective" is a global solution for you and that particular problem. .
Go with recent survey paper, select few algorithm and apply.. . :) . Also try to select the better multi - objective function for the problem. . finally its all about the combination of Selection of Objective function, Constraints, and initial Conditions . .
:)
S.K. Lakshmanaprabu
Thank-you for your suggestion it's useful to me.
It is necessary to understand different metaheuristics and how they work in order to be able to select one. Another important source is the previous literature on your problem, where you can find useful discussions and insights regarding different metaheuristic approaches and applications.
Please check the following links:
https://www.researchgate.net/publication/301725899_Vertical_Supply_Chain_Integrated_Decisions_A_Critical_Review_of_Recent_Literature_and_a_Future_Research_Perspective
www2.cscamm.umd.edu/publications/BookChapter_CS-09-13.pdf
Noha A. Mostafa
I express my gratitude for your time and suggestion.
Thanking you
You are quite lazy, Vivek. :-) I read in one response that your problem apparently is a multiobjective one. This is an interesting error - as the original question did NOT talk about a multiobjective problem, and it is quite telling: it's weird that there is such a focus on multiple objectives and metaheuristics in this community - or so it seems to me.
Metaheuristics cannot be trusted, as they have rather little theory behind them, and they do not easily adapt to a problem with a special structure and special properties. I've been here at RG for a rather long while, and I was hoping that what with all the posts and discussions, there would eventually be less metaheuristics left, and more scientific methodologies applied to the problem we face. Not so, and its' a shame.
Dr Professor Michael Patriksson
Greeting.
I wont comment on the first line of your response, that might be true. I must says yes i'm but at the same time i'm also ready to learn always. Since i am being very new to the metaheuristics i am learning facts still and i'm not vey sure how to make problems and use it for that.
Professor i have seen your several response to many questions on related topics. Your answer seems to be a fact for that problem. I have learnt manythings by reading it and also hope to see further.
Professor i'm also agree on each point you shared in response about metaheuristics especially this "Metaheuristics cannot be trusted, as they have rather little theory behind them, and they do not easily adapt to a problem with a special structure and special properties." Because they never promises a actual solution but near optimal.
Sir, i welcome your response.
Vivek: My response to you is this: Buy a few books on linear optimization and combinatorial optimization, and learn the real trade from them.
A good start in combinatorial optimization is Laurence Wolsey's short book Integer Programming (Wiley, 1998). There you will learn about good models versus bad models of a problem, for example, and through exercises you will also get many insights.
Dimitrii Bertsimas has a linear programming book (Introduction to Linear Programming) that is quite good (published by Athena).
This should keep you busy for a while, and help you orient yourself. :-)
Professor Dr. Michael Patriksson
Thank-you very much for suggesting books, definitely i'll buy and go through it. I express my gratitude for your time and suggestions it matters a lot to me .
•Meta-heuristics are a set of approximate algorithms (e.g. GA, SA, TS, MAS, ANN….) used to find optimal or near optimal solutions for a certain problem. The problem could be scheduling, logistic,… There is no guarantee that an algorithm or hybrid algorithm always able to get optimal solutions. You should survey different types of meta-heuristic approaches and decide what you can do.
•Check these papers:
•https://www.researchgate.net/publication/325274277_Developing_a_genetic-based_multi-objective_algorithm_to_optimise_job_shop_scheduling_problems
•https://www.researchgate.net/publication/315477712_Optimizing_Dynamic_Flexible_Job_Shop_Scheduling_Problem_Based_on_Genetic_Algorithm
I agree with the above answers.
I would suggest those few more points of interest :
- A first step would be to be sure that a metaheuristic is really the best option regarding your problem. Metaheuristic involve alea in order to get around certain issues in optimization problems (for exemple a high number of local optima). Be sure that the problem you're tackling possesses those issues (else a heuristic could be enough).
- The introduced alea has the effect of making metaheuristic "iteration consuming" (i.e. it generally takes a certain amount of iterations to get a correct result), meaning that your cost function should be as quick as possible to be computed (else the converging time could be crippling).
- Alea also means that for a given problem, one can get different response, this is not necesserally kindly regarded (yet) in the industry area.
- Finally, if metaheristics are what you go for, take a look at the homogeneity of your variables. If you have many different types of variables, some methods are more suited than others.
With that in mind, plus the above commentaries, you should be able to take a good decision.
I wish you the best of luck.
- Badges
- Science topic
- Similar topics
- Filing