Open AccessFace Recognition Based on CD-RBM and BM-ILM
Author(s) -
Weisong Qiao,
Binbin Chen,
Jiabao Zhao,
Guo Shu-dong
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1802/3/032077
Subject(s) - computer science , restricted boltzmann machine , artificial intelligence , boltzmann machine , pattern recognition (psychology) , facial recognition system , boosting (machine learning) , artificial neural network , classifier (uml) , feature extraction , gradient boosting , random forest
In recent years, the application of neural networks in face recognition has improved its accuracy greatly. However, it still suffers from the disappearance of the network gradient and the training cost. In this paper, we propose a new hybrid method which combines the contrast divergence algorithm (CD) accelerated Restricted Boltzmann Machine (RBM) and the Integrated Learning Method (ILM) with the Boosting algorithm (BM). First, we use RBM to establish a minimum energy model of the sample distribution. Then we use CD to speed up the feature extraction of samples in RBM. And then we build a multi-classifier with heterogeneous samples for ILM by BM. Finally, we experience our proposed method on the ORL database and the AR database. The results show that, with a simpler neural network determined by CD-RBM and BM-ILM, the recognition accuracy is better than that without the proposed method.