Multi-level algorithms for solving problems of parametric optimization based on bioinspired heuristics

Parametric optimization tasks are currently being used in various application areas. These tasks may include weather forecasting on meteo station, the calculation of the parameters of electric motors, search of weights coefficients in the neural network. This paper presents a hybrid bionic algorithm for solving the problems of parametric optimization. Also, it describes a series of experiments, which were confirmed by theoretical estimates, that identified the optimal parameters of the algorithm. The time complexity of the algorithm was $O(n^4)$, the value of the time offset, the quality of the solutions obtained via hybrid heuristics for a large number of input parameters are presented. Thus, in the course of the experiments, the number of input parameters for 100 or more a hybrid algorithm never got into a local optimum, and the solution found was approached or equal to the global.