zbMATH — the first resource for mathematics

Pseudo-optical neural networks – rectilinear models. (English. Russian original) Zbl 0930.68130
Autom. Remote Control 57, No. 12, Pt. 2, 1807-1814 (1996); translation from Avtom. Telemekh. 1996, No. 12, 145-154 (1996).
Summary: Consideration is given to networks of interfering neurons whose potentials vary as a result of interference of periodic inputs. Since the phase differences between signals depend on the lengths of links between neurons, the network geometry is important. Holographic effects in networks with rectilinear parallel neural layers are investigated, and a neural model of a lens is constructed.
68T05 Learning and adaptive systems in artificial intelligence