Optically addressed ferroelectric liquid-crystal devices present certa
in characteristics that make them attractive for all-optical implement
ations of neural architectures. This is particularly true for the impl
ementation of analog self-organizing systems such as Kohonen's self-or
ganizing map. The optical architecture proposed here is based on the p
rinciple of a resonator memory. The main novelty lies in the two-dimen
sional, continuous, binary information encoding and organization. Basi
c operations such as neural decision, weight updating, and information
spatial ordering are investigated. The principle of a learning proced
ure close to the Hebbian rule is discussed.