Which algorithm is better for optimal parameter selection ? Glowworm swarm optimization or particle swarm optimization or hybrid Genetic - PSO algorithm ?
I recommend the TLBO, it is very fast and do not request tuning parameters.
Article Teaching-Learning-Based Optimization: A novel method for con...
For me GSO is the Gothenburg Symphony Orchestra. I prefer that over any metaheuristics in general, and especially over PSO. ;-)
Seriously though, I would not take the route of implementing a meta-heuristic; it is not elegant, there is very little theory, and it can almost always be beaten by a hard-core mathematical optimization methodology, adapted to the problem. It seems to me to be a waste of resources to take that route.
The verdict is always to solve the problem such that you always identify an optimum solution - it may take more time, but you will also know that you have identified the very best - and I will always take that route.
PSO can converge quickly but always falls into premature problem, while GSO is easy to capture many peaks of multimodal function due to its dynamic sub-groups but also easy to be trapped in problems of low convergence and low precision. Combining the two algorithms can balance the diversity and convergence.
So don't use them, then! I fail to see why hordes of scientists - without a clear motivation than "I don't know the theory and practice of mathematical optimisation" - utilise inferior tools from the book of metaheuristics, when there are proven methods within mathematical optimisation. It's just dumb - and because of the fact that you cannot claim that your solutions are optimal, you may lose your clients, for example, as you cannot answer that the output is optimal. In the long run, your clients - if any - will go elsewhere to find optimal solutions. I have a hard time figuring out why you persist with those inferior tools.
No, it cannot - and it was not even intended to - solve a problem optimally - it's only a heuristic! This thread is full of such errors.
Heuristics are methods that use shortcuts to produce perhaps good-enough solutions given a limited time frame or deadline. Heuristics are a flexibility technique for quick decisions, particularly when working with complex data. ... Heuristic is derived from the Greek word meaning “to discover”.
See? "Good-enough" is the term used on the Wiki page. Good enough is NOT the same as the optimum, of course, and it's up to you if you can tolerate that the output from a heuristic is bad, which in the case of real-life problems definitely may be too far off from an optimal solution. A further bug with these methods is that you do not know how bad the output might be. You may be facing a solution that is hardly better than the one you started with!
PSO cannot stop at an optimum, even if you happen to start there - it's going to look elsewhere, and never come back. It's a lousy method, if you need to find an optimum. Stop using bad methods, and learn to use mathematical optimisation tools, instead.
- Badges
- Science topic
- Similar topics
- Mathematics