Embedded Files
Information is what generates information, say, evolution.
The Latest News: I joined South China Normal University (SCNU) as an Associate Professor!
2026.07.18 Our GECCO paper "A Learning-Based Cooperative Coevolution Framework for Heterogeneous Large-Scale Global Optimization" is selected as the Best Paper at GECCO 2026. Congrats for Qiu and other co-authors.
2026.06.17 One paper on RL-assisted multitask optimization got accepted by IEEE TEVC (SCI Q1 Top, IF=12.0).
2026.05.14 I am awarded as "Gold Reviewer" by ICML 2026.
2026.04.21 One paper got ACM GECCO 2026 "Best Paper Award Nomination", congrats!
2026.03.24 Five papers on MetaBBO got accepted by ACM GECCO 2026.
2026.03.07 Our Workshop on Learning-assisted Algorithm Design got accepted by IJCAI-ECAI 2026.
2026.01.15 One paper on LLM-based Optimization got accepted by IEEE TEVC (SCI Q1 Top, IF=12.0).
2025.12.29 Our work on ES for LLM's Reasoning got accepted by Bridge LMReasoning@AAAI 2026.
2025.12.24 The Chinese version of our survey paper got accepted by Chinese Jounal of Computers (计算机学报).
2025.11.08 One paper on automatic BBO benchmark generation got accepted by AAAI 2026.
2025.09.19 Two papers on MetaBBO (benchmark+algorithm) got accepted by NeurIPS 2025.
2025.09.13 Our Workshop on Learning-assisted Evolutionary Algorithm Design has been held successfully online.
2025.05.05 Our survey on MetaBBO got accepted by IEEE TEVC (SCI Q1 Top, IF=12.0).
G. Scholor: Zy. Ma
Email: mzy@ieee.org
Biography
Biography
Hi my friend, I am Zeyuan Ma, welcome to visit my homapage. I received my B.Eng. degree from School of Computer Science and Engineering, South China University of Technology (SCUT), Guangzhou, China, in 2022. In 2022-2026, I pursued my Ph.D. degree at SCUT, supervised by Prof. Yue-Jiao Gong. I joined AI4Opt group at Singapore Management University (SMU) as a visiting student in 2025, co-supervised by Prof. Yue-Jiao Gong and Prof. Zhiguang Cao. Currently, I am an Associate Professor at School of Artificial Intelligence, South China Normal University (SCNU). My research interests initially focus on learning-assisted optimization algorithms, where I establishsed a systematic paradigm (Meta-Black-Box Optimziation, MetaBBO) that automates algorithm designs in evolutionary optimization through meta learning. Currently, I extend my researches toward several cutting-edge topics such as evolutionary LLMs post-training, HyperAgents system with world model, LLMs for optimization and generalized optimization system in realworld applications. I have (co-) authored over 30 publications, most of which are published in top-tier conferences and journals such as ICML, ICLR, NeurIPS, AAAI, KDD and IEEE TEVC. I also actively serve as reviewer/PC member for many well-known AI conferences and journals including the mentinoed ones. I have participated in organizing LEAD 2025 , OPEA 2025 and IJCAI-LEAD 2026 workshops. More recently, I was invited as an Associate Editor of Journal of Intelligent and Sustainable Systems (JISS).
In the end of 2021 (approximately), my supervisor Prof. Yue-Jiao Gong, teamate Hongshu Guo and I started to establish a special task-force on Learning-assisted Optimization. Our initial intention is to explore how to use advanced learning techniques to enhance exsiting evolutionary optimizers in terms of their optimization performance and cross-domain generalization capability. Along this vision, we gathered many talented and hard-working student members, including undergraduate students, master students and a few PhD students. In the last four years, we have made certain progresses in developing diverse MetaBBO paradigms, analyzing the advantages and limitations behind and adapting them in the wild. Some of our papers serve as hightlighted contributions in evolutionary computation community, especially in automated algorithm design field. We believe works done by this team could promote the research edge of both Evolutionary Computation and Optimization.
Featured Works
Featured Works
My research efforts involve the whole life-cycle of learning-assisted automated algorithm design. I list following three core aspects.
[Survey] I have done a careful literature study on learning-assisted optimization papers and techniques. Such comprehensive understanding helps foster an universal perspective on Meta-Black-Box Optimization and hence results in a systematic survey on this field. This survey is entitled "Toward Automated Algorithm Design: A Survey and Practical Guide to Meta-Black-Box-Optimization", and is published at IEEE Transactions on Evolutionary Computation (SCI Q1 Top, IF=12.0).
[Benchmarking Platform] To help those who want to learn, develop or use the built-in working principles in learning-based optimization approaches, I spent three years on developing MetaBox benchmark platform (paper1, paper2, code, tutorial). The MetaBox papers are puiblished at NeurIPS conference (CCF A) and selected as Oral presentation. While there is still some improvement space for MetaBox, I suggest researchers and students, in particular those newcomers to learn how to use MetaBox for your project, since it really provides you a correct templated developing path.
[Algorithms] I have developed and published several key MetaBBO techniques that widely impacts the resarches in this field. They are accepted and included in CCF A conferences such as ICML, ICLR, NeurIPS, AAAI, KDD and SCI Q1 Top journals such as IEEE TEVC and TSMC. I list them at here.
Page updated
Google Sites
Report abuse