Nature of complex number and complex-valued neural networks

Abstract

We discuss the nature of complex number and its effect on complex-valued neural networks (CVNNs). After we review some examples of CVNN applications, we look back at the mathematical history to elucidate the features of complex number, in particular to confirm the importance of the phaseand-amplitude viewpoint for designing and constructing CVNNs to enhance the features. This viewpoint is essential in general to deal with waves such as electromagnetic wave and lightwave. Then, we point out that, although we represent a complex number as an ordered pair of real numbers for example, we can reduce ineffective degree of freedom in learning or self-organization in CVNNs to achieve better generalization characteristics. This merit is significantly useful not only for waverelated signal processing but also for general processing with frequency-domain treatment through Fourier transform.

Keywords

electromagnetic wave / lightwave / coherence / adaptive processing in sensing and imaging / learning logic / neural hardware

References

[1] Hirose A. Complex-Valued Neural Networks. Heidelberg: Springer-Verlag, 2006

[2] Hirose A. Complex-Valued Neural Networks: Theories and Applications. Singapore: World Scientific Publishing, 2003

[3] Hirose A. Complex-valued neural networks. IEEE Computational Intelligence Society (CIS) Video Archives, Tutorial, IJCNN 2009 Atlanta

[4] Hirose A. Applications of complex-valued neural networks to coherent optical computing using phase-sensitive detection scheme. Information Sciences-Applications, 1994, 2(2): 103-117

[5] Hirose A. Dynamics of fully complex-valued neural networks. Electronics Letters, 1992, 28(16): 1492-1494

[6] Hirose A. Continuous complex-valued back-propagation learning. Electronics Letters, 1992, 28(20): 1854-1855

[7] Birx D L, Pipenberg S J. A complex mapping network for phase sensitive classification. IEEE Transactions on Neural Networks, 1993, 4(1): 127-135

[8] Zhang Y, Ma Y. CGHA for principal component extraction in the complex domain. IEEE Transactions on Neural Networks, 1997, 8(5): 1031-1036

[9] Sawada H, Mukai R, Araki S, Makino S. Polar coordinate based nonlinear function for frequency-domain blind source separation. IEICE Transactions on Fundamentals of Electronics, Communications, and Computer Sciences, 2003, E86A(3): 590-596

[10] Hara T, Hirose A. Plastic mine detecting radar system using complex-valued selforganizing map that deals with multiple-frequency interferometric images. Neural Networks, 2004, 17(8-9): 1201-1210

[11] Hara T, Hirose A. Adaptive plastic-landmine visualizing radar system: effects of aperture synthesis and featurevector dimension reduction. IEICE Transactions on Electronics, 2005, E88-C(12): 2282-2288

[12] Yang C C, Bose N. Landmine detection and classification with complex-valued hybrid neural network using scattering parameters dataset. IEEE Transactions on Neural Networks, 2005, 16(3): 743-753

[13] Masuyama S, Hirose A. Walled LTSA array for rapid, high spatial resolution, and phase sensitive imaging to visualize plastic landmines. IEEE Transactions on Geoscience and Remote Sensing, 2007, 45(8): 2536-2543

[14] Masuyama S, Yasuda K, Hirose A. Multiple mode selection of walled-LTSA array elements for high resolution imaging to visualize antipersonnel plastic landmines. IEEE Geoscience and Remote Sensing Letters, 2008, 5(4): 745-749

[15] Nakano Y, Hirose A. Improvement of plastic landmine visualization performance by use of ing-CSOM and frequencydomain local correlation. IEICE Transactions on Electronics, 2009, E92-C(1): 102-108

[16] Yamaki R, Hirose A. Singular unit restoration in interferograms based on complex-valued Markov random field model for phase unwrapping. IEEE Geoscience and Remote Sensing Letters, 2009, 6(1): 18-22

[17] Suksmono A B, Hirose A. Adaptive complex-amplitude texture classifier that deals with both height and reflectance for interferometric SAR images. IEICE Transaction on Electronics, 2000, E83-C(12): 1905-1911

[18] Aizenberg I, Paliy D V, Zurada J M, Astola J T. Blur identification by multilayer neural network based on multivalued neurons. IEEE Transactions on Neural Networks, 2008, 19(5): 883-898

[19] Goh S, Mandic D P. Nonlinear adaptive prediction of complex valued non-stationary signals. IEEE Transactions on Signal Processing, 2005, 53(5): 1827-1836

[20] Goh S L, Mandic D P. An augmented extended Kalman filter algorithm for complex-valued recurrent neural networks. Neural Computation, 2007, 19(4): 1-17

[21] Hirose A, Eckmiller R. Behavior control of coherent-type neural networks by carrier-frequency modulation. IEEE Transactions on Neural Networks, 1996, 7(4): 1032-1034

[22] Suksmono A B, Hirose A. Beamforming of ultra-wideband pulses by a complex-valued spatio-temporal multilayer neural network. International Journal of Neural Systems, 2005, 15(1): 1-7

[23] Hirose A, Eckmiller R. Coherent optical neural networks that have optical-frequency-controlled behavior and generalization ability in the frequency domain. Applied Optics, 1996, 35(5): 836-843

[24] Kawata S, Hirose A. Frequency-multiplexed logic circuit based on a coherent optical neural network. Applied Optics, 2005, 44(19): 4053-4059

[25] Kawata S, Hirose A. Frequency-multiplexing ability of complex-valued Hebbian learning in logic gates. International Journal of Neural Systems, 2008, 12(1): 43-51

[26] Hirose A, Higo T, Tanizawa K. Efficient generation of holographic movies with frame interpolation using a coherent neural network. IEICE Electronics Express, 2006, 3(19): 417-423

[27] Tay C S, Tanizawa K, Hirose A. Error reduction in holographic movies using a hybrid learning method in coherent neural networks. Applied Optics, 2008, 47(28): 5221-5228

[28] Hirose A, Asano Y, Hamano T. Developmental learning with behavioral mode tuning by carrier-frequency modulation in coherent neural networks. IEEE Transactions on Neural Networks, 2006, 17(6): 1532-1543

[29] Lee D L. Improvements of complex-valued hopfield associative memory by using generalized projection rules. IEEE Transactions on Neural Networks, 2006, 17(5): 1341-1347

[30] Novey M, Adali T. Complex ICA by negentropy maximization. IEEE Transactions on Neural Networks, 2008, 19(4): 596-609

[31] Li H, Adali T., A class of complex ICA algorithms based on the kurtosis cost function. IEEE Transactions on Neural Networks, 2008, 19(3): 408-420

[32] Widrow B, McCool J, Ball M. The complex LMS algorithm. Proceedings of the IEEE, 1975, 63(4): 719-720

[33] Georgiou G M, Koutsougeras C. Complex domain backpropagation. IEEE Transactions on Circuits and Systems II, 1992, 39(5): 330-334

[34] Takeda M, Kishigami T. Complex neural fields with a hopfield-like energy function and an analogy to optical fields generated in phase-conjugate resonators. Journal of Optical Society of America A, 1992, 9(12): 2182-2191

[35] Ebbinghaus H D, Hermes H, Hirzebruch F, Koecher M, Mainzer K, Neukirch J, Prestel A, Remmert R. Numbers (Chapter 3, Section 2). Berlin: Springer-Verlag, 1983

[36] Copson E. An Introduction to the Theory of Functions of a Complex Variable. Oxford: Clarendon Press, 1935

[37] Kuroe Y, Taniguchi Y. Models of orthogonal type complexvalued dynamic associative memories and their performance comparison. In: Proceedings of International Conference on Artificial Neural Networks. 2007, 838-847

[38] Leung H, Haykin S. The complex backpropagation al gorithm. IEEE Transactions on Signal Processing, 1991, 39(9): 2101-2104

[39] Benvenuto N, Piazza F. On the complex backpropagation algorithm. IEEE Transactions on Signal Processing, 1992, 40(4): 967-969

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg