A parallel Bayesian optimization method based on adaptive surrogate model

被引：0
|
作者：
Lyu Z.-M. [1 ]
Wang L.-Q. [1 ]
Zhao J. [1 ]
Liu Y. [1 ]
机构：
[1] School of Control Science and Engineering, Dalian University of Technology, Dalian
来源：
Kongzhi yu Juece/Control and Decision | 2019年 / 34卷 / 05期
关键词：
Adaptive; Bayesian optimization; Expectation improvement; Parallel; Surrogate;
D O I：
10.13195/j.kzyjc.2017.1449
中图分类号：
学科分类号：
摘要：
A parallel Bayesian optimization method based on the adaptive surrogate model is proposed to solve the computationally expensive problems. Based on the multi-points expected improvement criterion, one can evaluate the batch of points simultaneously. For the problem that the global surrogate model is difficult to be constructed with a large number of historical data from the parallel optimization, an improved data parallel local learning methodology for Gaussian Process modeling is proposed to construct the local surrogate models online. Furthermore, in order to reduce the computational cost of the multi-points expected improvement criterion, a heuristic hierarchical optimization strategy is proposed to calculate the single point expected improvement criterion based on the adaptive surrogate model sequentially. Finally, the effectiveness of the proposed method is verified by 5 test problems. © 2019, Editorial Office of Control and Decision. All right reserved.
引用
收藏
页码：1025 / 1031
页数：6
相关论文
共 17 条
  • [1] Shahriari B., Swersky K., Wang Z., Et al., Taking the human out of the loop: A review of bayesian optimization, Proc of the IEEE, 104, 1, pp. 148-175, (2016)
  • [2] Marmin S., Chevalier C., Ginsbourger D., Differentiating the multipoint expected improvement for optimal batch design, Int Workshop on Machine Learning, Optimization and Big Data, pp. 37-48, (2015)
  • [3] Gonzalez J., Dai Z., Hennig P., Et al., Batch bayesian optimization via local penalization, Artificial Intelligence and Statistics, pp. 648-657, (2016)
  • [4] Haftka R.T., Villanueva D., Chaudhuri A., Parallel surrogate-assisted global optimization with expensive functions-Asurvey, Structural and Multidisciplinary Optimization, 54, 1, pp. 3-13, (2016)
  • [5] Schonlau M., Computer experiments and global optimization, pp. 67-79, (1997)
  • [6] Chevalier C., Ginsbourger D., Fast computation of the multi-points expected improvement with applications in batch selection, Int Conf on Learning and Intelligent Optimization, pp. 59-69, (2013)
  • [7] Ginsbourger D., Riche R.L., Carraro L., Kriging is well-suited to parallelize optimization, Computational Intelligence in Expensive Optimization Problems, pp. 131-162, (2010)
  • [8] Viana F.A.C., Haftka R.T., Surrogate-based optimization with parallel simulations using the probability of improvement, The 13th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conf., pp. 13-15, (2010)
  • [9] Desautels T., Krause A., Burdick J.W., Parallelizing exploration-exploitation tradeoffs in gaussian process bandit optimization, J of Machine Learning Research, 15, 1, pp. 3873-3923, (2014)
  • [10] Shah A., Ghahramani Z., Parallel predictive entropy search for batch global optimization of expensive objective functions, Int Conf on Neural Information Processing Systems. Montréal, 118, 19-20, pp. 3330-3338, (2015)
12