Тип публикации: доклад, тезисы доклада, статья из сборника материалов конференций
Конференция: International Scientific Conference Reshetnev Readings 2015; Красноярск, Россия; Красноярск, Россия
Год издания: 2016
Идентификатор DOI: 10.1088/1757-899X/122/1/012024
Ключевые слова: Algorithms, Continuous variables, Discrete variables, Global minimization, Mixed variables, Non-differentiable, Optimum solution, Relative value, Sampling points, Global optimization
Аннотация: The algorithms of global non-differentiable minimization of functions on set of the mixed variables: continuous and discrete with unordered specific possible values are constructed. The method of optimization is based on selective averaging of required variables, on adaptive reorganization of the sizes of admissible domain of trialПоказать полностьюmovements and on use of relative values for minimised functions. Existence of discrete variables leads to solution of a sequence of global minimization problems of the functions in space of only continuous variables at the presence: 1) of their inequality restrictions for each problem; 2) of the general inequality restrictions for all problems (i.e. at the absence of dependence of functions fore inequality restrictions from discrete variables). In the first case, presence of discrete variables with unordered non-numeric possible values leads to solution of sequence of problems of global minimization of multiextreme functions on set only of continuous variables at the presence of their inequality restrictions. As a result, among the received optimum solutions the best is selected. In the second variant all minimized functions is convoluted in each sampling point in one multiextreme function and this function is minimised on continuous variables. © Published under licence by IOP Publishing Ltd.
Журнал: IOP Conference Series: Materials Science and Engineering
Выпуск журнала: Vol. 122, Is. 1