On Very Large Scale Hamiltonian Neural Nets
Wiesław Sieńko, Wiesław Citko · 2003
This paper presents how to design very large scale lossless neural nets (LONN), which can be used as Walsh-spectrum analyzer. This analysis relies on the orthogonality of weight matrix W where W is Hurwitz-Radon matrix. The unique feature of the LONN is the possibility to treat them either as algorithms or as Hamiltonian physical objects (Walsh Transformation Processors).