A Semi-supervised Canonical Correlation Analysis Algorithm Based on Deep Boltzmann Machine
Journal Title: 河南科技大学学报（自然科学版） - Year 2016, Vol 37, Issue 2
Abstract
To solve the problem that the canonical correlation analysis( CCA) algorithm was not adapted to the high level of correlation,through pattern classification,an improved algorithm was proposed based on semisupervised canonical correlation analysis with deep Boltzmann machine( DBM). Firstly,the implicit feature and the explicit feature were extracted by using the deep Boltzmann machine. By using the annotated pairwise constraints of the sample information,the most effective identification trait was constructed. The simulation experiments were conducted via ORL,Yale and AR face databases. The results show that the proposed algorithms are better than other methods.
Authors and Affiliations
Wen JIANG, Lin QI
Keywords
Related Articles
- EP ID EP461021
- DOI 10.15926/j.cnki.issn1672-6871.2016.02.010
- Views 85
- Downloads 0
Improvement and Analysis of Oncolytic Virotherapy Kinetic Model
An improved kinetic model of tumor therapy with oncolytic virus was investigated. The boundedness and positivity of solutions were obtained by analyzing the model. The instability of zero equilibrium,the global stability...
Microwave Assisted Preparation of Zinc Cobaltite Nanoflowers and Lithium Storage Properties
Zinc cobaltite ( ZnCo2O4) nanoflowers were prepared by microwave assisted liquid phase method and annealing treatment. The effects of different annealing temperatures on performance of as-prepared samples were investigat...
Optimization of Excavation Method for Shallow Buried Tunnel Under-passing Highway with Large Pipe Curtain Construction
Based on the project of DK3 + 538 ～ DK3 + 588 tunnel section lower crossing G5001 high speed public road in Chongqing railway hub east ring road Xinbaiyangwan tunnel, the optimization of excavation method for large pipe...
Analysis of ZigBee Network Performance in Mine Water Inrush Application
Aiming at the reliability of data transmission in Zigbee wireless sensor network in the application of mine water inrush monitoring, the main controller STM32 F107 RB, radio frequency wireless transceiver CC2540 and sens...
Generalized Hermite Spectral Method of Heat Transfer on Whole Line
The generalized Hermite functions with scaling factor were used as basis function to expand numerical solution,which approximated the exact solution of heat transfer on the whole line. Algorithms scheme was constructed b...