Image recognition based on improved convolutional deep belief network model
- PDF / 1,361,534 Bytes
- 15 Pages / 439.37 x 666.142 pts Page_size
- 114 Downloads / 298 Views
Image recognition based on improved convolutional deep belief network model Wang Hongmei 1,2
& Liu Pengzhong
1,2
Received: 26 September 2018 / Revised: 8 July 2019 / Accepted: 30 September 2019 # Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract
Aiming at the homogeneity of convolution kernels in Convolutional Deep Belief Network (CDBN), a cross-entropy-based sparse penalty mechanism suitable for Convolutional Restricted Boltzmann Machine (CRBM) model is introduced which makes the hidden layer units of the whole network in a lower activation state. On this basis, a parameter learning algorithm is applied to compensate the gradient by introducing the prior information of the samples, which alternates the supervised learning and unsupervised learning. The experimental results show that the proposed model can weaken the homogeneity of convolution kernels and improve the supervising and predicting ability of the network. The recognition rate on simulated dataset achieves 97.45%, which is increased 5.12% and 1.29% than Convolutional Neural Network (CNN) and the traditional CDBN model, respectively. At the same time, test error rate on common dataset MNIST also shows that the proposed model is more effective than some other state-of-the-art deep learning models. Keywords Image recognition . Convolutional deep belief network . Convolutional restricted boltzmann machine . Homogeneity . Gradient diffusion
1 Introduction Image recognition is a very important research topic in computer vision [1, 4, 5]. The main challenge in image recognition is to extract the features that are able to promote the discrimination capability. High accuracy is difficult to be obtained by the traditional image recognition algorithms based on human-designed features. Contrary to the traditional feature extraction algorithm, deep learning (DL) extracts features by imitating the learning mechanism of the
* Wang Hongmei [email protected]
1
School of Astronautics, Northwestern Polytechnical University, Xi’an, China
2
National Key Laboratory of Aerospace Flight Dynamics, Northwestern Polytechnical University, Xi’an, China
Multimedia Tools and Applications
human brain. As a result, the deep learning model can mine and learn the inherent laws and patterns of data autonomously [23]. In addition, deep learning is a non-linear network structure with strong self-construction and generalization ability, and has been widely applied in the fields of speech recognition, image understanding and natural language processing as a result. The representative deep learning models include Restricted Boltzmann Machine (RBM), Deep Belief Network (DBN), CNN and so on [12]. DBN is one of the popular networks due to its effectiveness in unsupervised feature learning. RBM is the basic unit of DBN. The layer-by-layer pre-training results of RBM are used as the initial value of the BP network parameters in DBN. This mechanism can effectively prevent the network falling into the local extreme during the global fine-tune. However, DBN netwo
Data Loading...
