Please use this identifier to cite or link to this item:
Title: Genetic Algorithm For Convolutional Neural Network Hyperparameter Tuning
Authors: Santoso, Fian Yulio
Keywords: Auto-tuning Hyperparameter;Convolutional Neural Network;Genetic Algorithm;Harmony Search
Issue Date: 11-Apr-2023
Abstract: Image classification has been applied in various fields, from agriculture to health. In order to perform well, an image classification method needs an appropriate hyperparameter setting. However, tuning hyperparameters in a classification method is not an easy task. Many approaches have been developed to solve the problem, such as combining metaheuristics methods with a convolutional neural network (CNN). An example of metaheuristics method is genetic algorithm. Genetic algorithms have been proven to optimize machine learning and deep learning. This research contributes to the automatic tuning of hyperparameters usinggenetic algorithms. The proposed method is evaluated using MNIST dataset. The experiments results show that using a genetic algorithm for tuning hyperparameters automatically, the accuracy of validation data is 97.02% and the accuracy of training data is 99.77%. The performance of Genetic Algorithm is compared with Harmony Search. The accuracy of Harmony Search for validation dat is 83.96% and for training data is 88.33%. The Genetic algorithm takes 24,97 seconds for training while Harmony Search needs 56.44 seconds.
Appears in Collections:T2 - Master of Information Systems

Files in This Item:
File Description SizeFormat 
T2_972021002_Judul.pdf648.26 kBAdobe PDFView/Open
  Until 9999-01-01
330.32 kBAdobe PDFView/Open
T2_972021002_Daftar Pustaka.pdf298.81 kBAdobe PDFView/Open
T2_972021002_Formulir Lisensi dan Embargo.pdf
  Restricted Access
664.43 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.