Optimizations of Multi Optima Problems Using the Genetic Algorithm, Punch-Out Method, and Least Exponent Method

초록

The Genetic Algorithm (GA) is a very effective global search algorithm. When the dimension of the optimization problem becomes great, it is not a simple task for methods, such as gradient methods. For the accuracy of a given solution, the search frequency of the variables necessary for a variable axis is far greater with the dimension increase of the solution vector. When the smoothness of the problem is good, the small frequency of the variable will do, with the progress of the generations. However, when the problem has many sharp peaks, such as in data regression, several frequencies of the variable in each variable axis are needed to result in a tremendous total number of individuals. In order to resolve this kind of ineffectiveness for many variables with a moderate number of individuals, a punch-out method is proposed for effectively finding the optimum. The least square method is revisited here to suggest least exponent method to take care of the outlier problem in data regression. © 2022 Praise Worthy Prize S.r.l.-All rights reserved.

키워드