Adaptive momentum Levenberg-Marquardt RBF for face recognition

This paper investigates the application of Levenberg-Marquardt (LM) based radial basis function (RBF) neural networks for face recognition. The contribution of this paper is two-fold. First, we propose the use of Levenberg-Marquardt (LM) and adaptive momentum LM algorithm to update the weights and network parameters (centers and width). The purpose of the proposal of the latter algorithm is to further increase the learning efficiency of the RBF neural network. The second contribution of the paper is the adaptation of the high computational complexity LM-based RBF neural networks to the complex problem of face recognition. To reduce the computations required, dimension reduction was applied prior to the training of the networks. In addition to that, we have also proposed the use of prior knowledge to guess the initial values of the weights during initialization as oppose to random weights. The proposed methods were tested on the Yale database and were found to yield positive results that can further improve the learning efficiency of the networks for the application of face recognition.

[1]  Mohammad Bagher Menhaj,et al.  Training feedforward networks with the Marquardt algorithm , 1994, IEEE Trans. Neural Networks.

[2]  Adrian G. Bors,et al.  Minimal Topology for a Radial Basis Functions Neural Network for Pattern Classification , 1994 .

[3]  Byung-Joo Oh,et al.  Face Recognition using Radial Basis Function Network based on LDA , 2007, IEC.

[4]  Alex Pentland,et al.  Eigenfaces for Face Recognition , 1991 .

[5]  ErMeng Joo,et al.  Face recognition with radial basis function (RBF) neural networks , 2002 .

[6]  Nicolaos B. Karayiannis,et al.  Reformulated radial basis neural networks trained by gradient descent , 1999, IEEE Trans. Neural Networks.

[7]  Meng Joo Er,et al.  High-speed face recognition based on discrete cosine transform and RBF neural networks , 2005, IEEE Transactions on Neural Networks.

[8]  Stavros J. Perantonis,et al.  Two highly efficient second-order algorithms for training feedforward networks , 2002, IEEE Trans. Neural Networks.

[9]  Okyay Kaynak,et al.  Computing Gradient Vector and Jacobian Matrix in Arbitrarily Connected Neural Networks , 2008, IEEE Transactions on Industrial Electronics.

[10]  Meng Joo Er,et al.  Face recognition with radial basis function (RBF) neural networks , 2002, IEEE Trans. Neural Networks.

[11]  Chng Eng Siong,et al.  Gradient radial basis function networks for nonlinear and nonstationary time series prediction , 1996, IEEE Trans. Neural Networks.

[12]  Hao Yu,et al.  Improved Computation for Levenberg–Marquardt Training , 2010, IEEE Transactions on Neural Networks.

[13]  Kah Phooi Seng,et al.  Enhanced Audio-Visual Recognition System over Internet Protocol , 2009 .

[14]  Juyang Weng,et al.  Using Discriminant Eigenfeatures for Image Retrieval , 1996, IEEE Trans. Pattern Anal. Mach. Intell..

[15]  George W. Irwin,et al.  A New Jacobian Matrix for Optimal Learning of Single-Layer Neural Networks , 2008, IEEE Transactions on Neural Networks.

[16]  M. Turk,et al.  Eigenfaces for Recognition , 1991, Journal of Cognitive Neuroscience.

[17]  Stavros J. Perantonis,et al.  Levenberg-Marquardt algorithm with adaptive momentum for the efficient training of feedforward networks , 2000, Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks. IJCNN 2000. Neural Computing: New Challenges and Perspectives for the New Millennium.

[18]  Shang-Liang Chen,et al.  Orthogonal least squares learning algorithm for radial basis function networks , 1991, IEEE Trans. Neural Networks.