{"title":"Optimization of the Input Layer Structure for Feed-Forward Narx Neural Networks","authors":"Zongyan Li, Matt Best","volume":103,"journal":"International Journal of Electrical and Computer Engineering","pagesStart":673,"pagesEnd":679,"ISSN":"1307-6892","URL":"https:\/\/publications.waset.org\/pdf\/10001639","abstract":"This paper presents an optimization method for\r\nreducing the number of input channels and the complexity of the\r\nfeed-forward NARX neural network (NN) without compromising the\r\naccuracy of the NN model. By utilizing the correlation analysis\r\nmethod, the most significant regressors are selected to form the input\r\nlayer of the NN structure. An application of vehicle dynamic model\r\nidentification is also presented in this paper to demonstrate the\r\noptimization technique and the optimal input layer structure and the\r\noptimal number of neurons for the neural network is investigated.","references":"[1] D. Montana and L. Davis, \u201cTraining feedforward neural networks using\r\ngenetic algorithms\u201d. Proc.1989 International Joint Conf. Artificial\r\nIntelligence.\r\n[2] M. Khashei and M. Bijari, \u201cAn artificial neural network (p, d,q) model\r\nfor timeseries forecasting\u201d. Expert Systems with Applications Vol. 37\r\n,2010, pp. 479\u2013489\r\n[3] B. Pradhan and S. Lee and M. F. Buchroithner, \u201cA GIS-based backpropagation\r\nneural network model and its cross-application and\r\nvalidation for landslide susceptibility analyses\u201d. Computers,\r\nEnvironment and Urban Systems, Vol 34, 2010, pp. 216-235\r\n[4] M.A. Mohandes and S. Rehman and T.O. Halawani, \u201cA neural networks\r\napproach for wind speed prediction\u201d. Renewable Energy, Vol. 13,No. 3,\r\n1998, pp.345-354\r\n[5] L. Zhang and F. Tian and S. Liu etc, \u201cChaos based neural network\r\noptimization for concentration estimation of indoor air contaminants by\r\nan electronic nose\u201d. Sensors and Actuators A, Vol. 189, 2013, pp. 161-\r\n167.\r\n[6] P. G. Benardos, and G. C. Vosniakos, \u201cPrediction of surface roughness\r\nin CNC face milling using neural networks and Taguchi\u2019s design of\r\nexperiments\u201d Robotics and Computer Integrated Manufacturing, Vol.\r\n18, 2002, pp. 43\u2013354. [7] L. Ma, and K. Khorasani, \u201cA new strategy for adaptively constructing\r\nmultilayer feed-forward neural networks\u201d. Neurocomputing, 51, 2003,\r\npp. 361\u2013385.\r\n[8] J. P Ross, \u201cTaguchi techniques for quality engineering\u201d. New York:\r\nMcGraw-Hill, 1996.\r\n[9] S. D. Balkin and J. K. Ord, \u201cAutomatic neural network modeling for\r\nunivariate time series\u201d. International Journal of Forecasting, Vol. 16,\r\n2000, pp. 509\u2013515\r\n[10] M. M. Islam and K. Murase, \u201cA new algorithm to design compact two\r\nhidden layer artificial neural networks\u201d. Neural Networks, 14, 2001,\r\npp.1265\u20131278.\r\n[11] X. Jiang and A.H.K.S. Wah, \u201cConstructing and training feed-forward\r\nneural networks for pattern classification\u201d. Pattern Recognition Vol. 36,\r\n2003, pp.853\u2013867.\r\n[12] P. G. Benardos and G. C. Vosniakos, \u201cOptimizing feed-forward artificial\r\nneural network architecture\u201d. Engineering Applications of Artificial\r\nIntelligence, 20, 2007, pp. 365\u2013382.\r\n[13] G. Zhang and B. E. Patuwo and M. Y. Hu, \u201cForecasting with artificial\r\nneural networks: The state of the art\u201d, International Journal of\r\nForecasting, Vol. 14, Issue 1, March, 1998, pp 35\u201362.\r\n[14] D. Marquardt, \u201cAn Algorithm for Least-Squares Estimation of\r\nNonlinear Parameters\u201d. SIAM Journal on Applied Mathematics, Vol. 11,\r\nNo. 2, June 1963, pp. 431\u2013441.\r\n[15] M. T. Hagan and M. Menhaj, \u201cTraining feed-forward networks with the\r\nMarquardt algorithm\u201d, IEEE Transactions on Neural Networks, Vol. 5,\r\nNo. 6,1994, pp. 989\u2013993.\r\n[16] D. Whitley and T. Starkweather and C. Bogart, \u201cGenetic algorithms and\r\nneural networks: optimizing connections and connectivity\u201d. Parallel\r\nComputing Vol.14, 1990, pp. 347-361","publisher":"World Academy of Science, Engineering and Technology","index":"Open Science Index 103, 2015"}