Optimization when Cost of Optimization is Comparable to the Objective Function
Anirban Chaudhuri, Raphael T. Haftka · 2014
Anirban Chaudhuri, Raphael T. Haftka University of Florida, Gainesville, Florida, 32601 Abstract This work deals with the time critical nature of simulation-based design of engineering systems where delays in launching the product could lead to decreased market value. This work looks at two scenarios: (a) fixed function evaluation budget, and (b) cost included as a significant part of the objective function. For the first case, the implications of switching from an adaptive sampling strategy in surrogate-based optimization to a pure exploitation of surrogate prediction was investigated for test functions with competitive and non-competitive local optima. For the second case, an optimization formulation was proposed to include the cost in the objective function. In this case the cost affected not only the sampling process but also the stopping criterion for the algorithm.