An Algorithm for Finding the Global Extremum of a Partially Defined Function

The paper discusses the problem of finding the global minimum of a function that may be partially defined in the search domain. The objective function may not be defined within some search regions due to the nature of the optimized object and of the simulation method (for example, numerical instability of the simulation method used). In some cases, these regions are known; but in most cases, the information about them is missing. The paper gives a description of a global search algorithm for solving such class of problems. Numerical experiments confirming the efficiency of the proposed algorithm were carried out.