Hi,
Does anyone know any source (journal paper, conference paper, slides, etc.) that shows how to improve runtime of Genetic Algorithms (or any Evolutionary Algorithm)? I am hoping to find a way to increase the speed of GA to match or get as close as possible to the equivalent mathematical programming problem. I am looking for both algorithmic tweaks (e.g. giving the GA a good starting population) and implementation/technical tricks (e.g. using Cython if the GA code is written in Python).
Many thanks,
Majed
Have a look at https://link.springer.com/referenceworkentry/10.1007%2F978-3-540-92910-9_21
Article Genetic Algorithm – Survey Paper
Article Parallel genetic algorithms: A survey and problem state of the art
IMPACT OF GENETIC ALGORITHM OPERATORS ON ITS
PERFORMANCE IN SOLVING FLOW SHOP SCHEDULING PROBLEMS
Gamal M. Nawara, Adel A. Ibrahim, Raafat H. Elshaer, Hani A. Al–rawashdeh
Industrial Eng. Dept., Faculty of Eng., Zagazig University, Egypt
It is problem specific, but starting with a good set of solutions seems to be the way to go. You can also combine mathematical programming with GA, especially to generate good set of solutions. You just have to try a lot of different things. There is no guaranteed way to improve that would work on every problem instance.
Why focus on run-time? In many, many problem settings it is much more crucial to find the optimal solution.
And we all know that GA, ACO, and all the other meta-methods do not provide you with an optimal solution. Why settle with an inferior output?
Article Genetic Algorithm Parameter Optimization Using Taguchi Metho...
https://www.tandfonline.com/doi/abs/10.1080/00207543.2014.939244
Michael Patriksson , Thanks for your response.
I am dealing with a multi-objective problem and solving it with mathematical programming requires assigning weights to the different objective functions. I'd like to avoid using weights (because of the sensitivity and other known issues) and meta-methods seems like a feasible option to solve such problems.
Those who I'm solving this problem for, are more concern about long run-time of GA compared to equivalent mathematical programming method rather than the quality of optimality of final solutions. That's why I am keen to use GA but with similar speed to a mathematical programming counterpart.
You can develop a hybrid model with genetic algorithm to raise the performance of the genetic algorithm.
You may visit at https://doi.org/10.1007/s12652-020-02445-6
Isha Tyagi , I think this hybridisation adds to the overall run time rather than improving it. Although, Hybridising is good for improving the quality (optimality) of solutions found by an algorithm such as GA.
Am I correct?
This paper is easy to follow, written by two researchers applying GA in water resources management problems. I think it matches your interests and you'll find it useful.
Chapter Genetic Algorithms: A Mature Bio-inspired Optimization Techn...
Nikolaos Nagkoulis , thanks for sharing this paper with us which explains basics of a GA and provides some discussion around its pros and cons. However, I am looking for a way of making GA faster to make it as quick as a mathematical programming (ideally) for a same hypothetical problem.
Thanks
I hope you find the below publications useful :
two papers from Professor Deb (father of modern multi-objective evolutionary algorithms),
and one paper plus its code from professor Doerr .
https://ieeexplore.ieee.org/abstract/document/996017?casa_token=qpy4U0PTdVcAAAAA:8CjzRyXtnuXDcLnSNabaP4s20FR9XHQaV8kg-2vdIIw8KgN5Dn7OytERmpJRAaXW6KC5WCY6QQ
Conference Paper A Fast Elitist Non-dominated Sorting Genetic Algorithm for M...
Article Fast Genetic Algorithms
https://github.com/FastGA/fast-genetic-algorithms
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python