SEPARABLE MULTICRITERION DESIGN AND PERFORMANCE OPTIMIZATION OF AWMPs

An Aboriginal Whaling Management Procedure (AWMP) is an algorithm designed to operate on the results of an assessment (i.e. a statistical estimation problem), and its performance is tested through Monte Carlo simulation. I describe a method that separates the design, estimation, and simulation aspects of this problem. This allows algorithms created purely to emulate performance goals to be retrootted to meet design goals such as those in Subparagraph 13(a) of the Schedule. It allows evaluation of the design goals themselves. AWMPs can be tuned to enhance performance without a simulation-intensive search over the tuning parameter space. Finally, given thèwinning' AWMP and a collection of other AWMPs, one can construct a new AWMP that is at least as good as thèwinner.' This allows the best performance aspects of competing AWMPs to be combined. Ideally, and in examples shown, the newly constructed AWMP outperforms all previous candidates.