2

For which problems are Genetic Algorithms more suitable than Particle Swarm Optimization, and vice-versa? Are there any guidelines?

nbro
  • 39,006
  • 12
  • 98
  • 176
Student
  • 21
  • 2

2 Answers2

3

These kind of questions cannot be answered without looking at a particular project. Each algorithm has its particular strengths and weaknesses; and trade-offs in terms of use of resources (processing power and/or storage space, for example). If there was an objective answer, then the worse algorithm would surely fall in disuse.

It also depends what you mean by "better". Faster? Better score according to some evaluation measure? More robust (ie works with many diverse data sets)?

I would recommend looking at both algorithms in more detail, and trying to understand how they work. Then you should be able to find out which best fits your problem.

However, one problem with Particle Swarm Optimisation is that it is not well understood, so you might have to resort to trial-and-error.

Oliver Mason
  • 5,322
  • 12
  • 32
  • 1
    This should have been a comment (for the most part). You should have encouraged the asker to change his question to something like: when should I use genetic algorithms as opposed to PSO? – nbro May 06 '19 at 17:09
  • 1
    In order to save the original post (not sure if it is a good idea), I've reformulated the question, so that it's more "acceptable". You may want to review this answer to reflect those changes, as the question "when to use one over the other?" may be useful to some people. You may also consider the deletion of this answer, as it only partially addresses the current answer now (and, as I had said in a previous comment, this should probably have been a comment). – nbro Oct 14 '21 at 13:47
2

The paper Comparison between genetic algorithms and particle swarm optimization (1998, by Eberhart and Shi) does not really answer the question of when to use one over the other (this may be an open question), but at least it provides a comparison of how the methods work and what could affect their performance (i.e. which parameters or operators they use, and what the typical values are), so it may be worth reading it.

nbro
  • 39,006
  • 12
  • 98
  • 176