OPTIMIZATION OF HYPERPARAMETERS IN GENERATIVE ADVERSARIAL NETWORKS (GANs) USING GAUSSIAN AHP

The optimization of hyperparameters in Generative Adversarial Networks (GANs) using

the Gaussian analytical hierarchy process (Gaussian AHP), the focus of this study. The Combination

of machine learning techniques and multicriteria decision methods aims to improve performance

and the efficiency of GAN models. Using the Fashion MNIST dataset, the GAN models

are trained and the Gaussian AHP is applied to evaluate and optimize the hyperparameters based on

on multiple performance criteria, such as quality of images generated, training stability

and training time. The methodology is validated through iterative experiments,

in which hyperparameters are automatically adjusted based on the scores obtained, maximizing

the efficiency and quality of the model. The results indicate a significant improvement

imaging and computational efficiency.