Abstract. We describe and critique the convergence properties of filterbased evolutionary pattern search algorithms (F-EPSAs). F-EPSAs implicitly use a filter to perform a multi-objective search for constrained problems such that convergence can be guaranteed. We provide two examples that illustrate how F-EPSAs may generate limit points other than constrained stationary points. F-EPSAs are evolutionary pattern search methods that employ a finite set of search directions, and our examples illustrate how the choice of search directions impacts an F-EPSA’s search dynamics.
Lauren M. Clevenger, William E. Hart