Title: Evaluating the Swarm Metaphor based Heuristics for Continuous Optimization - Modern Performance Measures and the (Non-parametric) Statistical Perspectives
Dr. Swagatam Das (Electronics and Communication Sciences Unit of the Indian Statistical Institute, Kolkata, India)
Numerous metaheuristics have been proposed based on the life supporting activity of the swarming creatures in nature for solving optimization problems involving continuous search spaces. However, whether these algorithms will survive in the long run or whether their names will remain stipulated within the world of paper writing - this issue largely depends on the ease of implementation and efficiency in solving practical optimization problems. Most of the pioneering papers on swarm intelligence based metaheuristics generally exhibit the efficiency of the algorithm on a set of synthetic benchmark functions which are supposed to capture the different aspects of the real world optimization problems. This talk will focus on the benchmarking procedures adopted for comparing different metaheuristics meaningfully. The talk will elaborate on modern performance measures, design of experiments and provide some guidelines for selecting the suitable metaheuristic for solving a domain-specific problem. It will also discuss some recent non-parametric hypothesis test procedures to evaluate the comparative results. Finally some shortcomings of the modern statistical benchmarking procedures will be highlighted along with the discussion on some important future research directions.
Swagatam Das is currently serving as a faculty member at the Electronics and Communication Sciences Unit of the Indian Statistical Institute, Kolkata, India. His research interests include non-convex optimization and machine learning. Dr. Das has published one research monograph, one edited volume, and more than 300 research articles in peer-reviewed journals and international conferences. He is the founding co-editor-in-chief of Swarm and Evolutionary Computation, an international journal from Elsevier. He has also served as or is serving as the associate editors of several international journals of repute from IEEE, Elsevier and Springer. Dr. Das has 12000+ Google Scholar citations and an H-index of 54 till date. He has been associated with the international program committees and organizing committees of several regular international conferences including IEEE CEC, IEEE SSCI, and GECCO. He has acted as guest editors for special issues in journals like IEEE Transactions on Evolutionary Computation and IEEE Transactions on SMC, Part C. He is the recipient of the 2012 Young Engineer Award from the Indian National Academy of ngineering (INAE). He is also the recipient of the 2015 Thomson Reuters Research Excellence India Citation Award as the highest cited researcher from India in Engineering and Computer Science category between 2010 to 2014.
Title: Fitness Landscape Information for Better Optimization Search
Prof. Hideyuki TAKAGI (Kyushu University, Fukuoka, Japan)
We explain several approaches to obtain fitness landscape information and use it to enhance the performance of population-based optimization algorithms. Individuals of population-based optimization algorithms converge to the global optimum gradually with exchanging searching information among them. If we can know the global optimum area or better searching areas before individuals converge to the global optimum, its/their location information is helpful to accelerate optimization.
The first approach of obtaining the information is approximation of a fitness landscape and estimate the global optimum area from the hyper-surface of the approximated landscape. Unlike a surrogate model which is an approximated landscape for expensive optimization tasks, we must choose approximation approaches that provide us the estimated global optimum point. We introduce such approaches.
The second approach is the estimation of the convergence point of individuals from their moving directions mathematically. Suppose we define a moving vector as a directional vector from a parent to its offspring. Since individuals aim toward the global optimum according to search generations, we obtain many moving vectors toward to the global optimum and can expect that one point that is the nearest to these moving vector locates near the global optimum. Estimating the global optimum area before individuals converge on there must be helpful to accelerate optimization search.
The third approach is finding local optima areas of multimodal tasks. Individuals aims not only the global optimum but also local optima especially in early generations. We may be able to separate local areas using the approximated fitness landscape, i.e. the first approach, or using the directions of moving vectors aiming the different local optima. They can be new niche methods. The numbers of these local optima areas is directly related with the complexity of its fitness landscape, and we also discuss how to know the complexity of tasks, which is also helpful for optimization search.
高木英行(TAKAGI, Hideyuki) received the degrees of Bachelor and Master from Kyushu Institute of Design in 1979 and 1981, and the degree of Doctor of Engineering from Toyohashi University of Technology in 1991. He was a researcher at Panasonic Central Research labs in 1981 - 1995, was an Associate Professor of Kyushu Institute of Design in 1995 - 2003, and is a Professor of Kyushu University now. He was a visiting researcher at UC Berkeley in 1991-1993 hosted by Prof. L. A. Zadeh.
He had worked on neuro-fuzzy systems in 1987 - early 1990's and extended his interests to fusing neuro-fuzzy-genetic algorithms and human factors. Now, he aims Humanized Computational Intelligence and is focusing on interactive evolutionary computation (IEC) as a tool for this research direction and developing methods for enhancing evolutionary computation. The number of citations of the most cited his IEC paper is around 1,500 times, and his well cited papers can be found at Google Scholar Citations.
He has been a volunteer for IEEE Systems, Man, and Cybernetics (SMC) Society. Some of his contributions are: Vice President in 2006 - 2009: a member of Administrative Committee/Board of Governors in 2001 - 2010, and 2016 - 2018: Chair of SMC Japan Chapter in 2014 - 2017: Technical Committee (TC) Coordinator in 2004 - 2005: Chair of TC on Soft Computing in 1998 - 2004 and since 2008: Distinguished Lecturer in 2006 - 2011: Associate Editor of IEEE Transactions on SMC, Part B / Cybernetics since 2001.
See his further detail bio at his web page.
Title: Swarm Intelligence in Dynamic Environments
Prof. Shengxiang Yang (De Montfort University, Leicester, UK)
Swarm Intelligence (SI) represents the property that the collective behaviors of agents that interact locally with their environment cause coherent functional global patterns to emerge. SI algorithms are inspired from simple behaviors and self-organizing interaction among agents, such as ant foraging, bird flocking and fish schooling, and have been applied in different fields. Most SI algorithms have been developed to address stationary optimization problems. However, many real-world problems have a dynamic environment that changes over time due to many factors. For such dynamic optimization problems (DOPs), it is difficult for a conventional SI algorithm to track the changing optimum once changes occur. DOPs have attracted a growing interest from the SI community in recent years due to the importance in the real-world applications of SI algorithms. This talk will first briefly introduce the concepts of SI and DOPs, then review the enhancement strategies integrated into SI algorithms to address dynamic changes, and describe several detailed case studies on SI methods for DOPs. Finally, some conclusions will be made and the future work on SI in dynamic environments will be briefly discussed.
Shengxiang Yang is now a Professor of Computational Intelligence (CI) and the Director of the Centre for Computational Intelligence, De Montfort University, UK. He has worked extensively for 20 years in the areas of CI methods, including evolutionary computation, swarm intelligence and artificial neural networks, and their applications for real-world problems. He has over 240 publications in these domains. He has 6500+ Google Scholar citations and an H-index of 42. His work has been supported by UK research councils (e.g., Engineering and Physical Sciences Research Council (EPSRC), Royal Society, and Royal Academy of Engineering), EU FP7 and Horizon 2020, Chinese Ministry of Education, and industry partners (e.g., BT, Honda, Rail Safety and Standards Board, and Network Rail, etc.), with a total funding of over GBP2M, of which two EPSRC standard research projects have been focused on CI for DOPs. He serves as an Associate Editor or Editorial Board Member of seven international journals, including IEEE Transactions on Cybernetics, Evolutionary Computation, Information Sciences, and Soft Computing. He is the founding chair of the IEEE CI Society (CIS) Task Force on Intelligent Network Systems and the chair of the IEEE CIS Task Force on Evolutionary Computation in Dynamic and Uncertain Environments. He has organized over 40 workshops and special sessions on CI in dynamic and uncertain environments for several major international conferences. He is the founding co-chair of the IEEE Symposium on CI in Dynamic and Uncertain Environments. He has co-edited over ten books, proceedings, and journal special issues. He has been invited to give over 10 keynote speeches/tutorials at international conferences, and over 40 seminars in different countries.