Images source: https://towardsdatascience.com/

Efficient and Self-tuning Nature-inspired Algorithms

Amirhossein Rajabi; Efficient and Self-tuning Nature-inspired Algorithms

Nature-inspired computation is a framework for global optimization algorithms that mimic optimization in nature. Famous examples include evolutionary algorithms (EAs), which follows Darwin’s principle of survival of the fittest, ant colony optimization (ACO), which simulates the foraging behavior of ants, and particle swarm optimization (PSO), which is inspired by bird flocks.

 

Methods of nature-inspired computation are well established in numerous engineering disciplines. They are easy to implement, yield surprisingly good results and are therefore popular in many sectors, including engineering, biomedicine and finance. For example, they are used to plan production processes, optimize networks, construct work pieces. Only recently, nature-inspired computation was perceived as a class of algorithms and analyzed using methods known from the theory of algorithms.

 

Nature-inspired algorithms come with a big variety of parameters that have to be set for the algorithm to work satisfactorily. Knowing the best possible parameter setting in advance through theoretical research spares the practitioner the task of possibly time-consuming parameter tuning w.r.t. the underlying optimization algorithm and avoids the use of possibly inefficient rules of thumb or other predefined settings. Still, it is not always straightforward to apply the existing theoretical knowledge. For example, if one does not fully understand the structure of the optimization problem at hand and lacks a theoretical problem description, it is not clear which theoretical statement can be applied. Even more, usually there is no single parameter setting that is always optimal. Therefore, we investigate algorithms that automatically adjust their parameters during the run, so-called self-tuning nature-inspired algorithms.

 

The aim of this research project is to lay the foundations of a comprehensive theory of self-tuning nature-inspired optimization algorithms, building on the initial results from the literature and the methodological reservoir developed for more classical nature-inspired algorithms.

 

     

 

PhD project by Amirhossein Rajabi

Research section: Algorithms, Logic and Graphs

Principal supervisor: Carsten Witt

Co-supervisor: Paul Fischer

Title of project: Efficient and Self-tuning Nature-inspired Algorithms

Project start: 01/06/201931/05/2022

Contact

Amirhossein Rajabi
PhD student
DTU Compute
+45 52 61 09 45

Contact

Carsten Witt
Professor
DTU Compute
+45 45 25 37 22

Contact

Paul Fischer
Associate professor, head of section
DTU Compute
+45 45 25 37 13