Demand of New Type Augmentation Design from Practical Need

  • 报告人: 杨斯元博士
  • 报告人单位: 中芯国际集成电路新技术研发(上海)有限公司
  • 报告地址: 数学学院西303报告厅
  • 报告时间: 5月28日 10:20--11:00
  • 主题摘要:
Economic supersaturated Quadratic Model Designs have been important designs particually for every expensive experiments. They can be realized by either uniform design or Bayesian Modification of Optimal Design. Augmenting the supersaturated design by Optimal design does not add number of new runs arbitrarily and also these designs are model dependent. Sequential augmenting with Uniform design with quasi arbitrarily increased number of runs and with advantage of model free has been successfully realized by Professor Zhou Yongdao and his student recently. An example is illustrated in the paper to demonstrate the important advantage of such design and further demand of additional sequential design types are also brought up from practical need. The key aspects of the demanded new type designs include the consideration of blocking factor, using DSD design as the initial started supersaturated design, capability to augment original 3 level uniform design to 4 or more levels augmented design, augmenting with new center and/or new experimental ranges, and capability to augment a UD design mixed with both numerical variables and categorical variables. This presentation aims to bring up attention and interest from academic scholars to work out these new designs to meet challenging demand from practice with very expensive experiments.