Home > Research
Special Session on Experimental Analysis of Metaheuristics
OverviewIn Kenneth Sörensen's treatise "Metaheuristics--the metaphor exposed" [1], it is noted that "Papers on metaheuristics are not published if they contain an interesting insight, but if the methods they present are successful players in the up-the-wall game, and either perform better on average or beat some best-known result." We extend this observation with a specific example -- many new or modified techniques will claim to improve exploration to achieve a better result, but no actual measurements of exploration in either the new or original method are provided. In fact, a measurable definition of exploration may not have been used at all [2].
The lack of definition leads to a lack of measurement, which leads to a lack of experimentation in the analysis of metaheuristics. This Special Session will provide a forum for all components of this food chain, leading toward an improved understanding of how metaheuristics operate. We are particularly interested in novel experiments that provide quantitative measurements of specific features which can affect the performance of methods on benchmark functions or real-world problems.
Examples of important features which can benefit from new formal/measurable definitions
- Exploration and Exploitation
- Attraction Basin, Escape from Local Optimum
- Diversity, Search Trajectory
- Selection Pressure, Premature Convergence
Examples of experimental analysis to measure these important features
- Experiments to measure/observe specific features such as Exploration, Exploitation, Escape from Local Optimum, etc.
- Experiments to show the effect on performance based on measurements of a given feature
- Studies on the measurable differences of relevant techniques in terms of a given feature
- Analysis and/or experimentation on the effects that specific parameters or sub-components of metaheuristics can have on observed measurements for a given feature
- Experiments or studies (e.g. visualization) on the search trajectories and/or solution distributions generated by metaheuristics
- Analysis of search tractories in different types of search spaces (e.g. convex, deceptive, noisy, etc).
- Analysis of key characteristics of combinatorial and/or continuous domains.
[1] Sörensen, Kenneth. "Metaheuristics--the metaphor exposed." International Transactions in Operational Research 22.1 (2015): 3-18.
[2] Črepinšek, Matej, Shih-Hsi Liu, and Marjan Mernik. "Exploration and exploitation in evolutionary algorithms: A survey." ACM computing surveys (CSUR) 45.3 (2013): 1-33.
Organizers
Stephen Chen is an Associate Professor in the School of Information Technology at York University, Toronto, Canada. His research focuses on analyzing the mechanisms for selection, exploration, and exploitation in techniques designed for multi-modal optimization problems. He is particularly interested in the development and analysis of non-metaphor-based heuristic search techniques. He has 80 peer-reviewed publications including 30 CEC papers, and he has previously presented 3 tutorials and organized 2 workshops at CEC/WCCI conferences.
Alexandros Tzanetos is an Assistant Professor in the Department of Computing, School of Engineering, Jonkoping University, Jonkoping, Sweden, and Professeur Associé at the Université de Sherbrooke, Sherbrooke, Canada. His main areas of research interest are Evolutionary Computation, Artificial Intelligence, Operational Research, and applications in real-world optimization problems. He is the Section Editor of Artificial Intelligence for the academic journal Data in Brief. He is also a member of several associations and groups, such as the Special Interest Group on Genetic and Evolutionary Computation (SIGEVO), the Institute of Electrical and Electronics Engineers (IEEE), the EURO Working Group on Metaheuristics (EU/ME), the Canadian Operational Research Society (CORS), the Swedish Operations Research Association (SOAF), the Hellenic Artificial Intelligence Society (EETN), and the Swedish Artificial Intelligence Society (SAIS). Dr. Tzanetos is one of the two Management Committee members representing Sweden in the COST Action CA22137 - Randomised Optimisation Algorithms Research Network (ROAR-NET).
Jakub Kůdela is an Associate Professor at the Institute of Automation and Computer Science, Brno University of Technology, Brno, Czech Republic. His main research interests include developing computational methods for various optimization problems in engineering applications, benchmarking optimization algorithms, and investigating structural properties of optimization problems and methods.
Seyed Jalaleddin Mousavirad is a Senior Researcher at Mid Sweden University, Sundsvall, Sweden. Previously, he served as a Postdoctoral Research Fellow at the University of Beira Interior, Covilhã, Portugal, where he was actively involved in the European project called GreenStamp. His main research interests include evolutionary computation, deep and machine learning, image processing and computer vision, as well as measurement systems. He has published one book, eight book chapters, and more than 120 papers. He has organized Special Sessions at each of the last four CEC/WCCI conferences, at EvoStar, the International Conference on Neural Information Processing (ICONIP), and the International Work-Conference on Artificial Neural Networks (IWANN).
