Optimization of Hyper-parameter for CNN Model using Genetic Algorithm
Jihyun Yoo, Hyun-il Yoon, Hyeong-Gyun Kim, Hee-Seung Yoon, Seung-Soo Han · 2019
Recently CNN is not only widely used in the field of image recognition but also used in various fields such as classifying vibration data. Therefore, increasing the performance of CNN models is becoming more important. One of the various methods to improve the performance of CNN models is to optimize hyper-parameters. This paper presents a method for optimizing the hyper-parameters of CNN models that classify MNIST data using genetic algorithm. Population-based algorithms, different from previous studies, can be used to optimize several parameters at once. In addition, different types and ranges of parameters from the existing genetic algorithms are used. Using this method, the hyper-parameter values that best classify MNIST have been obtained and are presented.