This thesis proposes a hybrid adaptive learning algorithm called Particle Optimized Gradient Descent (POGD) to improve the convergence of convolutional neural networks (CNNs) in image classification. The proposed algorithm has a better protective effect from the local minimum and achieves a higher accuracy with fewer epochs. This research also emphasizes the importance of adjusting the parameters.
History
Pagination
p.
Open access
Yes
Language
eng
Degree type
Masters
Degree name
MEng
Copyright notice
All rights reserved
Editor/Contributor(s)
Sui Yang Khoo, Michael Norton
Faculty
Faculty of Science, Engineering and Built Environment