Learning Behaviors of Stochastic Gradient Radial Basis Function Network Algorithms for Odor Sensing Systems

ETRI Journal

关键词: least mean square (LMS);    stochastic gradient (SG);    center-gradient;    radial basis function network (RBFN);    Odor;   

Others  :  1185470
DOI  :  10.4218/etrij.06.0105.0046

【 摘 要 】

Learning behaviors of a radial basis function network (RBFN) using a singular value decomposition (SVD) and stochastic gradient (SG) algorithm, together named RBF-SVD-SG, for odor sensing systems are analyzed, and a fast training method is proposed. RBF input data is from a conducting polymer sensor array. It is revealed in this paper that the SG algorithm for the fine-tuning of centers and widths still shows ill-behaving learning results when a sufficiently small convergence coefficient is not used. Since the tuning of centers in RBFN plays a dominant role in the performance of RBFN odor sensing systems, our analysis is focused on the center-gradient variance of the RBFN-SVD-SG algorithm. We found analytically that the steady-state weight fluctuation and large values of a convergence coefficient can lead to an increase in variance of the center-gradient estimate. Based on this analysis, we propose to use the least mean square algorithm instead of SVD in adjusting the weight for stable steady-state weight behavior. Experimental results of the proposed algorithm have shown faster learning speed and better classification performance.

   

【 参考文献 】

• [1]P. Pelosi and K. Persaud, "Toward an Artificial Nose," Sensors and Sensory Systems for Advanced Robots, NATO ASI Series, vol. F43, 1988, pp. 49-70.
• [2]K. Persaud and P. Travers, Intelligent Instruments and Computers, 1991, p. 147.
• [3]K. Persaud and H. Byun, "Industrial Applications of Neural Networks," World Scientific, pp. 85-90.
• [4]H. Byun, K. Persaud, J. Kim, and D. Lee, "Improvement of RBF Network’s Performance for Application to Odour Sensing Systems," Proc. of ISOEN’99, 1999, pp. 237-240.
• [5]I. Cha and S.A. Kassam, "Interference Cancellation Using Radial Basis Function Networks," Signal Processing, vol. 47, 1995, pp. 247-268.
• [6]H. Byun, N. Kim, K. Persaud, J. Huh, and D. Lee, "Application of Adaptive RBF Networks to Odor Classification Using Conducting Polymer Sensor Array," Proc. of ISOEN’2000, 2000, pp. 121-126.
• [7]M.T. Vaki-Baghmished and N. Pavesi, "Training RBF Networks with Selective Backpropagation," Neurocomputing, vol. 62, 2004, pp. 39-64.
• [8]S.C. Fahlman, "An Empirical Study of Learning Speed in Backpropagation Networks," Technical Report CMU-CS-88-162, Carnegie Mellon University, Pittsburgh, PA 15213, Sept. 1988.
• [9]M.T. Vaki-Baghmished and N. Pavesi, "Backpropagation with Declining Learning Rate," Proc. of the 10th Electrotechnical and Computer Science Conf., Portoroz, Slovenia, vol. B, Sept. 2001, pp. 297-300.
• [10]I. Cha and S.A. Kassam, "Channel Equalization Using Adaptive Complex Radial Basis Function Networks," IEEE J. on Selected Areas in Comm., vol. 13. Jan. 1995, pp. 122-131.
• [11]B. Widrow and S.D. Stearns, Adaptive Signal Processing, Prentice-Hall, 1985.