Accelerating Evolutionary Algorithms to Solve High-Dimensional Expensive Problems via Autoencoders

A golden magnifying glass glass surrounded by formulas

Department of Electrical and Computer Engineering

Location: Burchard 714

Speaker: MengChu Zhou, Distinguished Professor, New Jersey Institute of Technology

ABSTRACT

High-dimensional computationally Expensive Problems (HEPs) in which a single fitness evaluation consumes hours or even days have attracted much attention from both academia and industry. Exponentially expanding search space and complex landscape brought by numerous decision variables make HEPs extremely challenging to be solved by traditional algorithms with limited physical/computational resources. Therefore, an Autoencoder-embedded Evolutionary Optimization (AEO) framework is invented to deal with them. To be specific, high-dimensional search space can be compressed to informative low-dimensional space by using an autoencoder as an effective dimension reduction tool. The search operation conducted in this low-dimensional space facilitates thSearch e population in convergence towards the optima. To balance the exploration and exploitation ability during optimization, two sub-populations are adopted to coevolve in a distributed/parallel fashion, wherein one is assisted by an autoencoder and the other undergoes a regular evolutionary process. Dynamic information exchange is conducted between them after each cycle to promote population diversity. Moreover, surrogate models can be incorporated into AEO (SAEO) to further boost its performance by reducing unnecessary fitness evaluation. Compared with the state-of-the-art algorithms for HEPs, AEO shows extraordinarily high efficiency for these challenging problems while SAEO can greatly improve the performance of AEO in most cases, thus opening new directions for various swarm optimization and evolutionary algorithms under both AEO and SAEO to tackle HEPs and greatly advancing the field of high-dimensional computationally expensive optimization. Their recent applications to human-cyber-physical systems and production scheduling are also illustrated.

BIOGRAPHY

Portrait of MengChu Zhou

MengChu Zhou received his Ph.D. in Computer and Systems Engineering from Rensselaer Polytechnic Institute in 1990. He then joined the New Jersey Institute of Technology (NJIT) and has been a Distinguished Professor in Electrical and Computer Engineering since 2013. Prof. Zhou’s research interests are in Petri nets, intelligent automation, AI, Cloud/edge Computing, the Internet of Things, big data, web services, and intelligent transportation. He has over 1,200 publications including 17 books, 850+ journal papers (650+ in IEEE transactions), 31 patents, and 32 book chapters. He is the Founding Editor of the IEEE Press Book Series on Systems Science and Engineering, and Associate Editor of the IEEE Internet of Things Journal, IEEE Transactions on Intelligent Transportation Systems, and IEEE Transactions on Systems, Man, and Cybernetics: Systems. He was Editor-in-Chief of the IEEE/CAA Journal of Automatica Sinica (2018-2022). He is a recipient of the Humboldt Research Award for US Senior Scientists from the Alexander von Humboldt Foundation, the Franklin V. Taylor Memorial Award and the Norbert Wiener Award from IEEE Systems, Man and Cybernetics Society, Excellence in Research Prize and Medal from NJIT, and Edison Patent Award from the Research & Development Council of New Jersey. He has been among the most highly cited scholars since 2012 and ranked top one in the field of engineering worldwide in 2012 by Web of Science. His present Google citation count is well over 72,400 with h-index being 137. He was ranked #99 in the world among the 2023 Top 1000 Scientists in Computer Science in the World by Research.com. He is a Fellow of IEEE, the International Federation of Automatic Control (IFAC), the American Association for the Advancement of Science (AAAS), the Chinese Association of Automation (CAA), and the National Academy of Inventors (NAI).