Науковий вісник НГУ

Optimizing feed-forward neural network weight based on orthogonal genetic algorithm

• 
• 

User Rating:  / 0

PoorBest 

Details

Category: Information technologies, systems analysis and administration

Last Updated on 02 April 2016

Published on 02 April 2016

Hits: 4759

Authors:

Xiaoqun Qin, Hunan International Economics University, Changsha, Hunan, China

Ye Chen, Hunan International Economics University, Changsha, Hunan, China

Abstract:

Purpose. Artificial neural network has been successfully applied in such fields as pattern recognition, intelligent control, combinatorial optimization and prediction. The integration of neural network and other traditional methods will promote the continuous development of artificial intelligence and information processing technology. The research combined Orthogonal Genetic Algorithm (OGA) and BP neural network. We have obtained an improved neural network-OGANet, which is helpful and effective for optimizing feed-forward neural network weight.

Methodology. In order to resolve the defects of the traditional BP network learning method: the tendency to be trapped in local extremum and low learning precision, we proposed a learning method to optimize BP neural network weight based on orthogonal genetic algorithm, which not only exerts the non-linear mapping capacity of BP neural network, but also strengthens the BP neural network learning ability.

Findings. The fundamental artificial neural network and orthogonal genetic algorithm were introduced. Then, we obtained the OGANet algorithm, and proved its effectiveness. The experimental result showed that the OGANet has high precision in learning, fast convergence speed and shows better performance than other neural network learning methods.

Originality. We made a study of optimizing feed-forward neural network weight based on orthogonal genetic algorithm. The analysis of the training results of the improved network has proved that the OGANet not only has fast training speed, but also, to certain extent, overcomes the tendency to be trapped in local minimum, which is a shortcoming of traditional BP neural network.

Practical value. The non-linear and self-adaptive information processing capacity of the neural network overcomes the defects of traditional artificial intelligence methods. This method effectively combines the advantages of genetic algorithm and BP neural network and it gives an excellent compromise between the global search of orthogonal genetic algorithm and the global development of BP network.  

References:

1. Pławiak, P., 2014. An estimation of the state of consumption of a positive displacement pump based on dynamic pressure or vibrations using neural networks. Neurocomputing, vol.144, no.20, pp. 471−483.

2. Renbo Luo, Wenzhi Liao and Youguo Pi., 2014. Discriminative supervised neighborhood preserving embedding feature extraction for hyperspectral image classification. TEL-KOMNIK Indonesian Journal of Electrical Engineering, vol. 12, no.6, pp. 4200−4205.

3. Buse Melis Ozyildirim and Mutlu Avci, 2014. Logarithmic learning for generalized classifier neural network. Neural Networks, vol.60, no.12, pp. 133−140.

4. Hrvoje Krstić, Željko Koški, Irena Ištoka Otković, Martina Španić, 2014. Application of Neural Networks in Predicting Airtightness of Residential Units. Energy and Buildings, vol. 84, no.12, pp. 160−168.

5. R. Krishnamoorthi and S. Sathiya Devi, 2013. A simple computational model for image retrieval with weighted multi-features based on orthogonal polynomials and genetic algorithm. Neurocomputing, vol.116, no.20, pp. 165−181.

6. Susmita Mall, S. Chakraverty, 2014. Chebyshev neural network based model for solving Lane–Emden type equations. Applied Mathematics and Computation, vol.247, no.15, pp. 100−114.

7. Tomás Rodríguez García, Nicoletta González Cancelas, Francisco Soler-Flores, 2014. The artificial neural networks to obtain port planning parameters. Procedia-Social and Behavioral Sciences, vol.162, no.19, pp. 168−177.

8. Donghong Zhao, 2014. Total variation differential equation with wavelet transform for image restoration. TELKOMNIKA Indonesian Journal of Electrical Engineering, vol.12, no.6, pp. 4747−4755.

9. Mohammad Khanzadeh and Moslem Malekshahi, Ali Rahmati, 2013. Optimization of loss in orthogonal bend waveguide: Genetic algorithm simulation. Alexandria Engineering Journal, vol.52, no.3, pp. 525−530.

10. Bernadete M.M. Neta, Gustavo H.D. Araújo, Frederico G. Guimarães, Renato C. Mesquita and Petr Ya. Ekel, 2012. A fuzzy genetic algorithm for automatic orthogonal graph drawing. Applied Soft Computing, vol.12, no.4, pp. 1379−1389.

2016_01_qin
2016-04-02  510.5 KB  942

• < Prev
• Next >