A comparison of feedforward and feedback simple cell models
There are two classes of networks that have been proposed as models for striate simple cells--feedforward and feedback networks. The goal of this thesis is to determine which type of network best accounts for the basic properties of simple cells. This problem was approached by generating computer models of simple cells, simulating the cortical response to visual test stimuli used in several key experiments, and comparing the simulation and experimental data. Our results demonstrate that virtually all simple cell properties are obtainable with either a feedforward network or a feedback network with weak isotropic inhibition. Models including a feedback mechanism that includes strong orientation-specific inhibition share many properties of biological simple cells but have receptive fields with anomalous spatial structure.
Blackman, David, "A comparison of feedforward and feedback simple cell models" (1992). Dissertations available from ProQuest. AAI9227616.