Abstract
An interchange optimization algorithm to construct partially-latinized designs is described. The objective function is a weighted linear combination of up to five functions, each of which corresponds to a blocking factor of the required design. Nested simulated annealing is used to address local optima problems. The average efficiency factors of the generated designs are assessed against theoretical upper bounds.
Original language | English |
---|---|
Pages (from-to) | 203-207 |
Number of pages | 5 |
Journal | Statistics and Computing |
Volume | 9 |
Issue number | 3 |
DOIs | |
Publication status | Published - 1999 |
Externally published | Yes |