Hello everyone,
I am looking to build genetic algorithms for production scheduling using python.
I found out that there are a few libraries like DEAP and PyEvolve and GeneAI and pyGAD , which support wide variety of genetic algorithms.
Alternatively, one can also try to code the algorithm up by themselves.
I'd like to know which approach is better?
Considerations include:
Continued code support for the program's usage lifetime, which is longer than the project duration
Flexibility and the ability to explore a wide range of crossover and mutation functions
Easiness of implementation ( avoiding premature convergence)
Code maintenance,
Looking forward to hearing from you all,
With warm regards,
Akhil Ramesh
GA itself is tested and proven, therefore both ways are fine unless until you want to modify any steps or want to build your own fitness function. In a later case, you have to code it from scratch.
I agree with the above comments; if your objective is to modify the GA code or hybrid it with another code to make a "memetic algorithm", then maybe writing code is good, but if you just want it to apply to solve your optimization problem, it is maybe unnecessary to write from scratch;
However, it is obviously very important that you have a good understanding of GA even in the latter case too
I suggest you first try to build you own codes from scratch that way you get to learn a lot like. And once you learn and understand how GA works then you can use library and stuff to optimize any problems.
Also writing GA from scratch is no that hard.. you can complete coding in like an hour or so but you need have clear idea on how GA works. May be the following vide on YouTube might be help full to start from scratch. It show how you can write your own GA code in 2 different platforms(MATLAB and Python). check it out
Link: https://www.youtube.com/watch?v=Fdk7ZKJHFcI&list=PLgH3sgdvgO4TQagjfi1FmPcdBjIlmk5s9
Write your own code from scratch then you will have more control and command.
I see that people recommends building the GA by yourself. If you would like a library where you will have high degree of flexibility and ease of use, then I recommend my library PyGAD: https://pygad.readthedocs.io
It has a wide range of parameters to customize the GA for different problems.
Ahmed Fawzy Gad
I am just curious. What about the maintenance of the library on a long-term basis.
Any research project would last only for a fixed duration and there may be portions of the tools developed which have to be commercialized within that, and their expected lifetime would be greater than that.
Would there be a problem of the library becoming obsolete?
Akhilnandh Ramesh
The library has some features that make it stand, at least at the current time.
I target supporting some new features in the future.
Commercializing the library is indeed good objective. I should go this way too!
As long as the genetic algorithm is still used in research, I think the library would still being active.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python