Uncertainty Modeling for Expensive Functions: A Rank Transformation Approach

An uncertainty model for an expensive function greatly improves the effectiveness of a design decision based on the use of a less accurate function. In this paper, we propose a method that exploits the concept of rank transformation to aggregate high fidelity information in a cost effective manner. This information is used to develop an empirical cumulative probability distribution function for residue such that it models the uncertainty with greater precision in the regions where the expensive function is potentially attractive as compared to other regions in the design space. The performance and robustness of the algorithm are demonstrated for univariate synthetic functions.

• Type: Journal
• Year: 2007
• Where: ENGL
• Authors: J. Umakant, K. Sudhakar, P. M. Mujumdar, C. Raghavendra Rao
• Keywords: Cost Effective Manner, Cumulative Probability Distribution Function, ENGL 2007, Expensive Function