1/2013 - 1 | View TOC | « Previous Article | Next Article » |
Automatic and Parallel Optimized Learning for Neural Networks performing MIMO ApplicationsFULGINEI, F. R. , LAUDANI, A. , SALVINI, A. , PARODI, M. |
Extra paper information in |
Click to see author's profile in SCOPUS, IEEE Xplore, Web of Science |
Download PDF (705 KB) | Citation | Downloads: 2,027 | Views: 7,611 |
Author keywords
neural networks, multivariate function decomposition, learning optimization, parallel computing, genetic algorithms
References keywords
neural(23), networks(14), network(9), salvini(6), riganti(6), fulginei(6), decomposition(5), problems(4), optimization(4), feed(4)
Blue keywords are present in both the references section and the paper title.
About this article
Date of Publication: 2013-02-28
Volume 13, Issue 1, Year 2013, On page(s): 3 - 12
ISSN: 1582-7445, e-ISSN: 1844-7600
Digital Object Identifier: 10.4316/AECE.2013.01001
Web of Science Accession Number: 000315768300001
SCOPUS ID: 84875323616
Abstract
An automatic and optimized approach based on multivariate functions decomposition is presented to face Multi-Input-Multi-Output (MIMO) applications by using Single-Input-Single-Output (SISO) feed-forward Neural Networks (NNs). Indeed, often the learning time and the computational costs are too large for an effective use of MIMO NNs. Since performing a MISO neural model by starting from a single MIMO NN is frequently adopted in literature, the proposed method introduces three other steps: 1) a further decomposition; 2) a learning optimization; 3) a parallel training to speed up the process. Starting from a MISO NN, a collection of SISO NNs can be obtained by means a multi-dimensional Single Value Decomposition (SVD). Then, a general approach for the learning optimization of SISO NNs is applied. It is based on the observation that the performances of SISO NNs improve in terms of generalization and robustness against noise under suitable learning conditions. Thus, each SISO NN is trained and optimized by using limited training data that allow a significant decrease of computational costs. Moreover, a parallel architecture can be easily implemented. Consequently, the presented approach allows to perform an automatic conversion of MIMO NN into a collection of parallel-optimized SISO NNs. Experimental results will be suitably shown. |
References | | | Cited By «-- Click to see who has cited this paper |
[1] K. H. Lim, K. P. Seng, L. M. Ang, S. W. Chin., "Lyapunov Theory-Based Multilayered Neural Network", IEEE Trans. on Circ. and Syst.-II: Express Briefs, vol. 56, no. 4, April (2009), pp. 305-309. [CrossRef] [Web of Science Times Cited 27] [SCOPUS Times Cited 33] [2] B. Yalcin, K. Ohnishi, "Infinite-Mode Networks for Motion Control", IEEE Transactions on Ind. Elect., vol. 56, no. 8, August (2009), pp. 2933-2944. [CrossRef] [Web of Science Times Cited 12] [SCOPUS Times Cited 15] [3] G. Capizzi, S. Coco, C. Giuffrida, A. Laudani, "A neural network approach for the differentiation of numerical solutions of 3-D electromagnetic problems", IEEE Trans. on Magnetics, Vol. 40, No. 2, pp. 953-956, 2004. [CrossRef] [Web of Science Times Cited 15] [SCOPUS Times Cited 19] [4] N. Morariu, S. Vlad, "Using Pattern Classification and Recognition Techniques for Diagnostic and Prediction," Advances in Electrical and Computer Engineering, vol. 7, no. 1, pp. 63-67, 2007. [CrossRef] [Full Text] [Web of Science Times Cited 2] [5] N. S. Thomaidis, G. D. Dounias, "A Hybrid Neural Network-Based Trading System", Proceeding on HAIS '09 Proceedings of the 4th International Conference on Hybrid Artificial Intelligence Systems, Pages 694 - 701 [CrossRef] [SCOPUS Times Cited 1] [6] P. Shuang, Y. Wei-kang and G. Mei ling "Application of Simulated Annealing BP Neural Network in Financial Crisis Early Warning", International Conference on Computational Intelligence and Software Engineering (CiSE), 2010, pp. 1-3, [CrossRef] [SCOPUS Times Cited 2] [7] L. O. Fedorovici, R. E. Precup, F. Dragan, R. C. David and C. Purcaru, "Embedding Gravitational Search Algorithms in Convolutional Neural Networks for OCR applications", 7th IEEE International Symposium on Applied Computational Intelligence and Informatics (SACI), 2012, pp. 125 - 130 [CrossRef] [SCOPUS Times Cited 21] [8] L. Jianyo, L. Yongchun, B. Jianpeng, S. Xiaoyun, L. Aihua. Flaw, "Identification Based on Layered Multi-subnet Neural Networks", Proceedings of Second Int. Conf. on Intelligent Networks and Intelligent Systems. 1-3 November (2009). Tianjin, China, pp.118-128. [CrossRef] [SCOPUS Times Cited 1] [9] A. Sun, A. Zhang, Y. Wang., "Largescale Artificial Neural Network Owning Function Subnets", Proceedings of 2006 IEEE Int. Conf. on Mechat. and Autom. June 25 - 28, (2006), Luoyang, China, pp. 2465-2470. [CrossRef] [SCOPUS Times Cited 2] [10] W. Haikun, D. Weiming, X. Sixin, "Designing Neural Networks Based on Structure Decomposition", Proceedings of the 3d World Congress on Intel. Cont. and Aut. June 28-July 2, (2000), Hefei, P.R. China. pp. 821-825. [CrossRef] [11] H. Kabir, Y. Wang, M. Yu and Q. J. Zhang, "High-Dimensional Neural-Network Technique and Applications to Microwave Filter Modeling", IEEE Trans on Micr. Theory and Tech., vol. 58, no. 1, January (2010), pp. 145-156. [CrossRef] [Web of Science Times Cited 74] [SCOPUS Times Cited 85] [12] F. Riganti Fulginei, A. Salvini, "Neural Network Approach for Modelling Hysteretic Magnetic Materials under Distorted Excitations", IEEE Transactions On Magnetics, vol. 48, p. 307 -310, 2012. [CrossRef] [Web of Science Times Cited 50] [SCOPUS Times Cited 51] [13] F. Riganti Fulginei, A. Salvini, C. Coltelli, "A Neuro-Genetic and Time-Frequency Approach Macromodeling Dynamic Hysteresis in Harmonic Regime", IEEE Transactions On Magnetics, vol. 39, p. 1401-1404, 2003. [CrossRef] [Web of Science Times Cited 14] [SCOPUS Times Cited 24] [14] S. Fiori, Singular Value Decomposition Learning On Double Stiefel Manifold, Int.Journal of Neural Sys., Vol. 13, No. 2 (2003) World Sci. Pub.Comp. pp. 1-16. [PubMed] [15] H. Trung Huynh, Y. Won, "Training Single Hidden Layer Feedforward Neural Networks by Singular Value Decomposition", Proceedings of 2009 Fourth Int. Conf. on Comp. Scie. and Conv. Inf. Tech. 30 November -2 December Seoul, Korea. pp. 1300-1304. [CrossRef] [SCOPUS Times Cited 15] [16] K. Rohani, M.S. Chen, M. T. Manry, "Neural Subnet Design by Direct Polynomial Mapping", IEEE Trans. on Neural Net., vol. 3, no. 6, November (1992). pp. 1024-1026. [CrossRef] [Web of Science Times Cited 14] [SCOPUS Times Cited 16] [17] F. Riganti Fulginei, A. Salvini, M. Parodi, "Learning Optimization Of Neural Networks Used For Mimo Applications Based On Multivariate Functions Decomposition". Inverse Problems In Science & Engineering, vol. 20, p. 29-39, 2012. [CrossRef] [Web of Science Times Cited 23] [SCOPUS Times Cited 28] [18] B. Curry, "Neural networks and seasonality: Some technical considerations", European Journal of Operational Research vol. 179, pp. 267-274, 2007. [CrossRef] [Web of Science Times Cited 11] [SCOPUS Times Cited 14] [19] E. J. Teoh, K. C. Tan, C. Xiang, "Estimating the number of hidden neurons in a feedforward network using the singular value decomposition", IEEE Trans. on Neural Networks, vol. 17, no. 6, pp. 1623-1629, nov. 2006. [CrossRef] [Web of Science Times Cited 94] [SCOPUS Times Cited 126] [20] G. Bebis, M. Georgiopoulos, "Optimal feed-forward neural network architectures", IEEE Potentials, October 1994, pp. 27-31. [CrossRef] [SCOPUS Times Cited 429] [21] C. Cernazanu, "Training Neural Networks Using Input Data Characteristics," Advances in Electrical and Computer Engineering, vol. 8, no. 2, pp. 65-70, 2008. [CrossRef] [Full Text] [Web of Science Times Cited 6] [SCOPUS Times Cited 7] [22] R. Mirsu, S. Micut, C. Caleanu, D. B. Mirsu, "Optimized Simulation Framework for Spiking Neural Networks using GPU's," Advances in Electrical and Computer Engineering, vol. 12, no. 2, pp. 61-68, 2012. [CrossRef] [Full Text] [Web of Science Times Cited 4] [SCOPUS Times Cited 4] [23] R. Reed, "Pruning algorithms-A review," IEEE Trans. Neural Networks, vol. 4, pp. 740-747, 1993. [CrossRef] [Web of Science Times Cited 1070] [SCOPUS Times Cited 1294] [24] T. Kwok, D. Yeung, "Constructive algorithms for structure learning in feedforward neural networks for regression problems," IEEE Trans. Neural Networks, vol. 8, n. 3, pp. 630-645, May 1997. [CrossRef] [Web of Science Times Cited 348] [SCOPUS Times Cited 414] [25] Wolpert D. H., Macready, W. G., "No free lunch theorems for optimization," IEEE Trans. on Evolutionary Computation, vol. 1, pp. 67-83, 1997. [CrossRef] [SCOPUS Times Cited 10229] [26] F. Riganti Fulginei, A. Salvini, "Comparative Analysis between Modern Heuristics and Hybrid Algorithms", COMPEL, The International Journal for Computation and Mathematics in Electrical and Electronics Engineering, MCB University Press, vol. 26, no. 2, pp. 264-273, March 2007. [CrossRef] [Web of Science Times Cited 32] [SCOPUS Times Cited 35] [27] F. Riganti Fulginei, A. Salvini, "Hysteresis model identification by the Flock-of-Starlings Optimization", International Journal Of Applied Electromagnetics And Mechanics, vol. 30, p. 321-331, 2009. [CrossRef] [Web of Science Times Cited 22] [SCOPUS Times Cited 27] [28] F. Riganti Fulginei, A. Salvini and G. Pulcini, "Metric-topological-evolutionary optimization". Inverse Problems in Science & Engineering (IPSE), vol. 20, p. 41-58, 2012 [CrossRef] [Web of Science Times Cited 22] [SCOPUS Times Cited 31] Web of Science® Citations for all references: 1,840 TCR SCOPUS® Citations for all references: 12,923 TCR Web of Science® Average Citations per reference: 63 ACR SCOPUS® Average Citations per reference: 446 ACR TCR = Total Citations for References / ACR = Average Citations per Reference We introduced in 2010 - for the first time in scientific publishing, the term "References Weight", as a quantitative indication of the quality ... Read more Citations for references updated on 2024-11-12 05:44 in 183 seconds. Note1: Web of Science® is a registered trademark of Clarivate Analytics. Note2: SCOPUS® is a registered trademark of Elsevier B.V. Disclaimer: All queries to the respective databases were made by using the DOI record of every reference (where available). Due to technical problems beyond our control, the information is not always accurate. Please use the CrossRef link to visit the respective publisher site. |
Faculty of Electrical Engineering and Computer Science
Stefan cel Mare University of Suceava, Romania
All rights reserved: Advances in Electrical and Computer Engineering is a registered trademark of the Stefan cel Mare University of Suceava. No part of this publication may be reproduced, stored in a retrieval system, photocopied, recorded or archived, without the written permission from the Editor. When authors submit their papers for publication, they agree that the copyright for their article be transferred to the Faculty of Electrical Engineering and Computer Science, Stefan cel Mare University of Suceava, Romania, if and only if the articles are accepted for publication. The copyright covers the exclusive rights to reproduce and distribute the article, including reprints and translations.
Permission for other use: The copyright owner's consent does not extend to copying for general distribution, for promotion, for creating new works, or for resale. Specific written permission must be obtained from the Editor for such copying. Direct linking to files hosted on this website is strictly prohibited.
Disclaimer: Whilst every effort is made by the publishers and editorial board to see that no inaccurate or misleading data, opinions or statements appear in this journal, they wish to make it clear that all information and opinions formulated in the articles, as well as linguistic accuracy, are the sole responsibility of the author.