Analogies between some learning laws widely used in self-organizing ne
ural networks and the dynamics of hologram creation in photorefractive
materials are outlined. Some possible optical implementations of opti
cal neuron units with photorefractive crystals are proposed. These sys
tems present the key characteristics of the adaptive units required fo
r the realization of self-organizing networks. The characteristics, su
ch as feature extraction in a set of images, are experimentally demons
trated.