Hiper Parametre Optimizasyonu Hyper Parameter Optimization

Abstract

The optimization of hyper parameters, which are the parameters that must be entered when designing machine learning models, has a positive effect on the performance of machine learning models as it reduces the operating cost. There are four methods (grid search, random search, Bayesian, and evolutionary algorithms) often used in the literature in this field. Within the scope of the study, these methods were discussed and their advantages and disadvantages were determined. When we look at the studies, it is noted that a hyper-parameter Analysis section, in which hyper-parameter values are selected at certain intervals and the connection between these values is analyzed, should be in all studies. © 2019 IEEE.