Associate Professor, Associate Head of Department Huazhong University of Science and Technology (HUST) IEEE Senior Member, chenghehust [at] gmail.com Dr. Cheng He is currently an Associate Professor with the School of Electrical and Electronic Engineering, Huazhong University of Science and Technology, China. His main research interests are Artificial/Computational Intelligence (including evolutionary multi-objective optimization, model-based optimization, large-scale optimization, etc.).
We have categorized our articles relating to surrogate-assisted optimization, large-scale optimization, multi-/many-objective optimization, and deep learning and its applications. You are welcome to cite these papers via the Bibtex.
- 11/2024: Thanks for Linqiang Pan's invitation from 2024 Operational Research Academic Symposium. I was invited to give a keynote speech on "Challenges and Opportunities of Evolutionary Large-Scale Multi-Objective Optimization for IntelliSense in Electric Power Measurement" is about the challenges and opportunities of large-scale multiobjective optimization. poster
- 08/2024: Thanks for the interview invitation from Blue Tech Wave during the DOCS 2024. This interview is about the challenges and opportunities of large-scale multiobjective optimization, and the interview is listed for your reference.
- 08/2024: Our paper titled "Unsupervised Online Poor Contact Detection in the Secondary Circuit of Voltage Transformers" is accepted to Measurement! The concept of ``knee point'' is embedded in unsupervised anomaly detection of the secondary circuit of voltage transformers.
- 07/2024: Our paper titled " Knowledge-assisted Differential Evolution based Non-contact Voltage Measurement for Multiconductor Systems in Smart Grid" is accepted to Advanced Engineering Informatics! Thanks to PhD Candidate Chaojun Ma for his hard work on the paper! This is my initial attempt at the embedding of physical knowledge in the design of evolutionary algorithms. More advanced embedding techniques are highly desirable! Let's see how we can make it more versatile and applicable.
- 03/2024: Our paper titled " Computationally Expensive High-dimensional Multiobjective Optimization via Surrogate-assisted Reformulation and Decomposition" is accepted to IEEE Transactions on Evolutionary Computation! Thanks to PhD Candidate Jianqing Lin for his hardworking on the paper! This work designed an SAEA that considers the balance of convergence, diversity, and computational complexity via problem reformulation and decomposition to handle high-dimensional EMOPs. The proposed TP-SAEA obtains well-converged and diverse solutions with only 509 real function evaluations, examined on six real-world instances with up to 12,000 decision variables. -->
|
Cheng He , Lianghao Li, Ye Tian, Xingyi Zhang, Ran Cheng, Yaochu Jin, Xing Yao IEEE Transactions on Evolutionary Computation, 2019 paper code poster LSMOF is an effective framework for large-scale multiobjective optimization. Generally, this framework can reduce the number of decision variables from 1000 to less than 20. The cost of FEs is less than 100,000 for conventional LSMOPs. |
|
Cheng He , Shuhua Huang, Ran Cheng, Tan Kay Chen Yaochu Jin, IEEE Transactions on Cybernetics, 2021 paper code poster GMOEA focuses on efficient offspring generation via learning from the distribution of promising solutions. GMOEA is capable of handling MOPs with up to 200 decision variables effectively, which is a new research direction for model-based evolutionary computation. |
|
Cheng He , Ran Cheng, Chuanji Zhang, Ye Tian, Qin Chen Xin Yao, IEEE Transactions on Evolutionary Computation, 2020 paper code TREE is a large-scale multiobjective optimization test suite extracted from the power dilivery system, aiming at handling real-time ratio error estimation of voltage transformers. Generally, the maximum number of decision variables is up to half a million and it includes constraints, which provides a guidance for the desgin of meaningful evolutionary algorithm. |
|
Shihua Huang, Zhichao Lu, Ran Cheng, Cheng He ICCV, 2021 arXiv code FaPN a simple yet effective top-down pyramidal architecture to generate multi-scale features for dense image prediction. It improves FPN's AP / mIoU by 1.5 - 2.6% on all tasks. |