Weijie ZHENG

2019-11-06

Weijie Zheng received Bachelor Degree (July 2013) in Mathematics and Applied Mathematics from Harbin Institute of Technology, and Doctoral Degree (October 2018) in Computer Science and Technology from Tsinghua University (Supervisor: Prof. Guangwen Yang). He was a visiting scholar (from April 2019 to May 2019) in Prof. Xin Yao’s group in the Department of Computer Science and Engineering at Southern University of Science and Technology. From May 2019, he is a postdoctoral researcher in Prof. Xin Yao’s group. Before that, he had an internship in parallel optimization in National Supercomputing Center in Wuxi.

 

His current research majorly focuses on the theoretical analysis and design of evolutionary algorithms. Comparing with the wide applications of evolutionary algorithms, the theoretical research falls behind. He devotes his effort to the theory analysis on evolutionary algorithms, and hopes that with the theoretical analysis, especially the runtime analysis, of evolutionary algorithms, researchers and practitioners could better understand the working principles, advantages and drawbacks of these black-box optimization algorithms so that they could design efficient algorithms for practical usage. He has published the relevant papers, such as in Theoretical Computer Science, and in International Conference on Parallel Problem Solving from Nature (PPSN), Genetic and Evolutionary Computation Conference (GECCO), and so on. He has a Best Paper Nomination at GECCO 2018, and serves as a Program Committee member at Theory Track of GECCO 2019.

 

His previous research also focused on parallel optimization / high-performance computing, especially on the Sunway TaihuLight Supercomputer. His co-authors and him have published several papers in this topic, like in ACM Transactions on Architecture and Code Optimization, and in IEEE International Parallel and Distributed Processing Symposium (IPDPS), International Conference for High Performance Computing, Networking, Storage and Analysis (SC), and so on.

Journal Papers

  • Doerr B*, Zheng W*. Working principles of binary differential evolution[J]. Theoretical Computer Science, 2019.
  • Zhao W, Fu H, Fang J, Zheng W, Gan L, Yang G. Optimizing convolutional neural networks on the sunway taihulight supercomputer[J]. ACM Transactions on Architecture and Code Optimization (TACO), 2018, 15(1): 13.
  • Zheng W, Yang G, Doerr B. Working principles of binary differential evolution[C]//Proceedings of the Genetic and Evolutionary Computation Conference. ACM, 2018: 1103-1110.
  • Fang J, Fu H, Zhao W, Chen B, Zheng W, Yang G. swdnn: A library for accelerating deep learning applications on sunway taihulight[C]//2017 IEEE International Parallel and Distributed Processing Symposium (IPDPS). IEEE, 2017: 615-624.
  • Zheng W, Fu H, Yang G. Tade: tight adaptive differential evolution[C]//International Conference on Parallel Problem Solving from Nature (PPSN). Springer, 2016: 113-122.

Research

在演化计算方法的理论分析上,其首次对二进制差分进化算法的运行机制建立了理论分析。其证明不同于其他演化计算方法,二进制差分进化算法是稳定的,即在长时间内,中性位的采样频率维持在初始值1/2附近。然而,稳定性也使得二进制差分进化算法很难搜索到对目标函数影响较小的变量的最优解,例如对于简单对称的OneMax问题,二进制差分进化算法需要问题规模指数量级的运行时间寻找到最优解。另一方面,其证明二进制差分进化算法可以快速优化最重要的决策变量。例如,主导位在种群规模对数量级的时间内收敛到最优值。