Discussing the article: "Population optimization algorithms: Resistance to getting stuck in local extrema (Part I)"
Interesting research!
What came to mind for some reason was dividing the search space into 4/9/16/... parts and running the algorithm on each subspace (but with fewer iterations) and then selecting the best result.
Andrey Khatimlianskii #:
That's an interesting study!
What came to mind for some reason was dividing the search space into 4/9/16/... parts and running the algorithm on each subspace (but with fewer iterations) and then selecting the best result.
It's great to see that the study has generated interest from readers.
Yes, dividing the space into zones, exploring the zones separately and then analysing the results makes practical sense.
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Check out the new article: Population optimization algorithms: Resistance to getting stuck in local extrema (Part I).
This article presents a unique experiment that aims to examine the behavior of population optimization algorithms in the context of their ability to efficiently escape local minima when population diversity is low and reach global maxima. Working in this direction will provide further insight into which specific algorithms can successfully continue their search using coordinates set by the user as a starting point, and what factors influence their success.
My thoughts and ideas leading to this research are the result of deep immersion in the topic and passion for scientific research. I believe that this work may become an important contribution to the field of algorithmic optimization attracting the attention of researchers and practitioners.
In this experiment, I propose to conduct a test aimed at assessing the resistance of algorithms to getting stuck in local extrema and, instead of randomly placing agents at the first iteration throughout the entire field of the search space, place them in the global minimum. The objective of the experiment is to search for a global maximum.
In such a scenario, where all the search agents of the algorithm are located at one point, we are faced with an interesting phenomenon - a degenerate population. This is like a moment of freezing time, where diversity in the population is reduced to a minimum. Although this scenario is artificial, it allows us to obtain interesting conclusions and evaluate the impact of reducing diversity in the population on the outcome. The algorithm should be able to get out of such a bottleneck and achieve a global maximum.
In this kind of stress test for optimization algorithms, we can reveal the secrets of agent interaction, their cooperation or competition, and understand how these factors affect the speed of achieving the optimum. Such analysis opens new horizons in understanding the importance of diversity in a population for the efficient operation of algorithms and allows us to develop strategies for maintaining this diversity to achieve better results.
To carry out the experiment, we need to first initialize the coordinates of the agents forcibly outside the algorithm, using the coordinates of the global minimum, before measuring the fitness function at the first epoch.
Such an experiment will allow us to evaluate resistance to extremely difficult conditions and the ability to overcome limitations.
Author: Andrey Dik