The RF-LISSOM model demonstrates that the afferent and lateral connections in the primary visual cortex can self-organize simultaneously and synergetically based on a single Hebbian adaptation process. The model is well supported by anatomical and physiological evidence for high contrast stimuli. It has already been very successful at modeling structural development and adult plasticity on the cortex.
The self-organization process stores long-range activity correlations between units into the lateral connections. During visual processing, this information is used to eliminate redundancies and to form an efficient sparse coding of the input. Training inputs can come from the visual environment or from internally generated sources. If the inputs have features that are oriented (i.e., are not radially symmetrical), the model will self-organize to represent that aspect of the inputs. Chapter 4 will examine how this self-organization leads to the development of orientation preferences. Later chapters will show that the same decorrelating process that forms a sparse coding of orientation and position also results in psychological artifacts known as the tilt aftereffect.