Implementasi Metode Optimasi Gradient Centralization untuk Pembuatan Model Klasifikasi Citra Pemandangan Alam
Abstract
Optimization algorithms are algorithms that are needed to properly train Neural Networks. Optimization algorithms help improve model performance by modifying the attributes of the neural network, such as weights and learning rate to further enchant the model. Gradient Centralization is a new optimization algorithm that optimizes by centralizing gradient vectors to have zero mean. This paper focuses on finding the optimal learning rate for Gradient Centralization and uses that learning rate to create a classification model to classify natural scene images. The optimal learning rate obtained by this research is 2e-5 and the model obtained 84,17% mean recall, 84,39% mean precision, and overall 83,60% accuracy.
References
[2] O'Shea, Keiron., & Nash, Ryan. (2015). An Introduction to Convolutional Neural Networks.
[3] R. Poojary and A. Pai, "Comparative Study of Model Optimization Techniques in Fine-Tuned CNN Models," 2019 International Conference on Electrical and Computing Technologies and Applications (ICECTA), 2019, pp. 1-4, doi: 10.1109/ICECTA48151.2019.8959681.
[4] Ondieki, B. (2015). Convolutional Neural Networks for Scene Recognition.
This work is licensed under a Creative Commons Attribution 4.0 International License.
The Authors submitting a manuscript do so on the understanding that if accepted for publication, the copyright of the article shall be assigned to JNATIA (Jurnal Nasional Teknologi Informasi dan Aplikasinya) as the publisher of the journal. Copyright encompasses exclusive rights to reproduce and deliver the article in all forms and media, as well as translations. The reproduction of any part of this journal (printed or online) will be allowed only with written permission from JNATIA (Jurnal Nasional Teknologi Informasi dan Aplikasinya). The Editorial Board of JNATIA (Jurnal Nasional Teknologi Informasi dan Aplikasinya) makes every effort to ensure that no wrong or misleading data, opinions, or statements be published in the journal.
This work is licensed under a Creative Commons Attribution 4.0 International License.