Abstract
Existing space-filling designs are based on the assumption that the experimental region is rectangular, while in practice this assumption can be violated. Motivated by a data centre thermal management study, a class of probability-based Latin hypercube designs is proposed to accommodate a specific type of irregular region. A heuristic algorithm is proposed to search efficiently for optimal designs. Unbiased estimators are proposed, their variances are given and their performances are compared empirically. The proposed method is applied to obtain an optimal sensor placement plan to monitor and study the thermal distribution in a data centre.
Original language | English (US) |
---|---|
Pages (from-to) | 961-968 |
Number of pages | 8 |
Journal | Biometrika |
Volume | 97 |
Issue number | 4 |
DOIs | |
State | Published - Dec 2010 |
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- General Mathematics
- Agricultural and Biological Sciences (miscellaneous)
- General Agricultural and Biological Sciences
- Statistics, Probability and Uncertainty
- Applied Mathematics
Keywords
- Computer experiment
- Heuristic search
- Latin hypercube design
- Sampling
- Space-filling design