S. Buchholz, K. Tachibana, E. Hitzer
Introduction to Theory, Construction and Application of Clifford Neural Networks
Tutorial Notes for 14th International Conf. on Neural Information Processing (ICONIP 2007), Japan, 13 Nov. 2007, 14 pages.

Introduction: Clifford neural networks is a handle for neural networks in the Clifford algebra domain. Clifford algebras are named after the mathematician William Kingdon Clifford. They subsume, for example, the real numbers, the complex numbers and the quaternions. The latter being a four-dimensional algebra invented byWilliam Rowan Hamilton. In the 1960s David Hestenes started to study Clifford algebras as a universal language for geometry for which he coined the term "Geometric algebra". A term that actually dates back to Clifford himself. Its geometric nature makes Clifford algebra a very interesting framework for neural computation. These notes are intented to sketch some fundamentals of the theory of Clifford neural computation. Fortunately, complex-valued neural networks are a vital topic themself. The recent book [11] provides an overview about their theory and applications. Quaternionic neural networks are an emerging topic. Their theory started with the pioneering work of Arena et al. [1].

[1] P. Arena, L. Fortuna, G. Muscato, and M. G. Xibilia. Neural Networks in Multidimensional Domains. Number 234 in LNCIS. Springer, New York, 1998.
[11] A. Hirose. Complex-Valued Neural Networks. Number 32 in Studies in Computational Intelligence. Springer, New York, 2006.

[ PDF ] 1074K

[Geometric Calculus Japan index]