Recurrent probabilistic dynamics: Applications to face recognition

摘要

The biological motivation for this work is to investigate the role of retinal image processing and the information content encoded in ganglia, LGN and higher levels of processing in the visual system that contribute to face recognition through combination of top-down and bottom-up recurrent dynamics. We propose a network model composed of three sub-structures that are probabilistic and dynamic with a time resolution for individual neuronal processes in the order comparable to the refractory period in spiking.