Abstract:
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.
Keywords:bio-inspired algorithm, multi-level algorithm, the ant algorithm, parameter optimization,
neural network.