Abstract
Oja’s principal subspace algorithm is a well-known and powerful technique for learning and tracking principal information of time series. However, Oja’s algorithm is divergent when performing the task of minor subspace analysis. In the present paper, we transform Oja’s algorithm into a dual learning algorithm in the sense of fulfilling principal subspace analysis as well as minor subspace analysis via geodesic search on Stiefel manifold. Also inherent stability is guaranteed for the proposed geodesic based algorithm due to the fact the weight matrix rigourously evolves on the compact Stiefel manifold. The effectiveness of the proposed algorithm is further verified in the section of numerical simulation.
Similar content being viewed by others
References
Comon P, Golub GH (1990) Tracking a few extreme singular values and vectors in signal processing. Proc IEEE 78(8):1327–1343
Li N, Guo G, Chen L, Chen S (2013) Optimal subspace classification method for complex data. Int J Mach Learn Cybern 4(2):163–171
Guo G, Chen S, Chen L (2012) Soft subspace clustering with an improved feature weight self-adjustment mechanism. Int J Mach Learn Cybern 3(1):39–49
Li J, Han G, Wen J, Gao X (2011) Robust tensor subspace learning for anomaly detection. Int J Mach Learn Cybern 2(2):89–98
Golub GH, Van Loan CF (1996) Matrix computations, 3rd edn. Johns Hopkins University Press, Baltimore
Juang BH, Kung SY, Kamm CA (1991) In: Proceedings of the 1991 IEEE workshop neural networks for signal processing. Princeton
Oja E (1992) Principal components, minor components, and linear neural networks. Neural Netw 5(6):927–935
Kung S-Y, Diamantaras KI, Taur J-S (1994) Adaptive principal component extraction (APEX) and applications. IEEE Trans Signal Process 42(5):1202–1217
Xu L (1993) Least mean square error reconstruction principle for self-organizing neural-nets. Neural Netw 6(5):627–648
Chen T, Amari S-I, Lin Q (1998) A unified algorithm for principal and minor components extraction. Neural Netw 11:385–390
Manton JH, Helmke U, Mareels IM (2005) A dual purpose principal and minor component flow. Syst Control Lett 54:759–769
Chen T, Amari S-i (2001) Unified stabilization approach to principal and minor components extraction algorithms. Neural Netw 14:1377–1387
Fiori S (2002) A minor subspace algorithm based on neural Stiefel dynamics. Int J Neural Syst 12(5):1–18
Kong X, Hu C, Han C (2012) A dual purpose principal and minor subspace gradient flow. IEEE Trans Signal Process 60(1):197–210
Abed-Meraim K, Attallah S, Chkeif A, Hua Y (2000) Orthogonal Oja algorithm. IEEE Signal Process Lett 7(5):116–119
Helmke U, Moore JB (1993) Optimization and dynamical systems. Springer, Berlin, Heidelberg
Yan W-Y, Helmke U, Moore JB (1994) Global analysis of oja’s flow for neural networks. IEEE Trans Neural Netw 5(5):674–683
Douglas SC, Kung S, Amari S-i (1998) A self-stabilized minor subspace rule. IEEE Signal Process Lett 5(12):328–330
Absil P-A, Mahony R, Sepulchre R (2009) Optimization algorithms on matrix manifolds. Princeton University Press, Princeton
Yoshizawa S, Helmke U, Starkov K (2001) Convergence analysis for principal component flows. Appl Math Comut Sci 11(1):223–236
Baldi P, Hornik K (1991) Backpropagation and unsupervised learning in linear networks. In: Chauvin Y, Rumelhart DE (eds) Backpropagation: theory, architectures and applications. Erlbaum Associates, NJ
Edelman A, Arias T, Smith S (1998) The geometry of algorithms with orthogonality constraints. SIAM J Matrix Anal Appl 20(2):303–353
Nishimori Y, Akaho S (2005) Learning algorithms utilizing quasi-geodesic flows on the Stiefel manifold. Neurocomputing 67:106–135
Abrudan TE, Eriksson J, Koivunen V (2008) Steepest descent algorithms for optimization under unitary matrix constraint. IEEE Trans Signal Process 56(3):1134–1147
Webers C, Manton JH (2006) A generalisation of the oja subspace flow. 17th international symposium on mathematical theory of networks and systems, pp 24–28
Acknowledgments
The authors would like to acknowledge the reviewers for their helpful comments and suggestions for improvements of this paper. Supported by the National Natural Science Foundation of China under Grant No.61002039 and No.61202254; The Fundamental Research Funds for the Central Universities under Grant No. DC12010216 and No. 0913130475.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Liu, L., Ge, R., Meng, J. et al. Dual subspace learning via geodesic search on Stiefel manifold. Int. J. Mach. Learn. & Cyber. 5, 753–759 (2014). https://doi.org/10.1007/s13042-013-0217-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13042-013-0217-x