AGN and CNN based optimizing method and device of single-objective optimization problem

Abstract

The invention discloses an AGN and CNN based optimizing method and device of a single-objective optimization problem. A trained AGN coding part is used to convert an optimization problem model in which the amount of decision variables is k (k is not 2) into an optimization problem model in which the amount of decision variables is 2, the two converted decision variables are input to a trained CNNtwo-decision-variable optimizing device to output a global optimal result rapidly, a trained AGN decoding part is used to convert the two decision variables in the global optimal result output by theCNN two-decision-variable optimizing device into an original space in which the amount of decision variables is k, and a final global optimal result of the optimization problem is output. Via the technical schemes provided by the invention, optimizing calculation of the CNN two-decision-variable optimizing device is linear parallel calculation, the optimizing speed is high, global optimization canbe searched directly, and the condition of local optimization is avoided. In addition, the method and device are also suitable for constrained optimization problems.

Description

technical field [0001] The invention relates to the technical field of computer applications, in particular to an AGN and CNN-based single-objective optimization problem-finding method and device. Background technique [0002] With the rapid development of computer application technology, more and more attention has been paid to the research on optimization problems. There are a large number of optimization problems in the real world, and the solution to the optimization problem can be divided into two steps. The first step is to convert a given actual optimization problem into an optimization problem with a canonical function expression through mathematical modeling, that is, to define the decision space, the target space and the mapping relationship between them; the second The first step is to obtain the optimal solution through calculation on the basis of the optimization problem expressed by the canonical function obtained in the previous step. [0003] Optimization p...

AI Technical Summary

Problems solved by technology

[0006] Traditional numerical optimization methods, such as Newton's method, tend to fall into the local optimum of the problem; while generalized evolutionary algorithms (including evolutionary algorithms and swarm intelligence algorithms) design iterations by simulating the behavior characteristics of biological groups or the development and structural characteristics of things. formula, the global search ability is strong, but it takes a long time to solve complex optimization problems and the accuracy of the approximate optimal solution is not enough. If the model parameters are not designed properly, the approximate optimal solution cannot be solved. In addition, for Constrained optimization problems, the difficulty of model design will be greatly increased and the model parameters will be difficult to determine, making the solution effect poor and difficult to meet the requirements of practical problems

Images