1. Haar Wavelet Neural Network Model
- Author
-
İbrahim Yücedağ, Uğur Güvenç, Hamdi Tolga Kahraman, Tuba Pala, Yusuf Sönmez, and [Belirlenecek]
- Subjects
Structure (mathematical logic) ,Artificial neural network ,Computer science ,business.industry ,Deep learning ,010102 general mathematics ,Convolutional neural network (CNN) ,Pattern recognition ,01 natural sciences ,Convolutional neural network ,Haar wavelet ,Haar Wavelet Transfrom ,Benchmark (computing) ,Artificial intelligence ,0101 mathematics ,business ,MNIST database - Abstract
2018 International Conference on Artificial Intelligence and Data Processing, IDAP 2018 -- 28 September 2018 through 30 September 2018 -- -- 144523 Convolutional neural networks, one of the most important methods of deep learning which is a popular and modern research topic. Nowadays, thismethod has been applied many problems in a short time and obtained successful results for science and the industry. The multi-layer structure adopted in the design of the convolutional neural network increases the network depth and thus leads to significant problems. In this study, Haar Wavelet Transform-based neural network structure is proposed. Proposed model reduces complexity and number of layers in the network structure. Performance ratios of the proposed model and the conventional model were tested on benchmark MNIST dataset. As a result, when the proposed Haar Wavelet Neural Network model and the convolutional neural network model are compared the accuracy is increased and running time is 6.5 times faster. © 2018 IEEE. 2-s2.0-85062483616
- Published
- 2019