Although much progress has been made in understanding the development and function of the visual cortex experimentally, the exact mechanisms are still not completely understood. With the advent of parallel supercomputers in the past few years, it has become viable to test theories about these mechanisms in detailed computational simulations. In the PIs previous project, a model of the primary visual cortex was developed that showed how the observed receptive fields, columnar organization, and lateral connectivity can arise through input-driven Hebbian self-organization. The current project continues the research in four directions: (1) The self-organizing model is extended to a hierarchy of visual representations, taking into account feedback from higher levels of visual processing. (2) Internal pattern generators are included in the model and the hypothesis that they implement genetic control over the self-organized structures is verified against data on chick imprinting and the development of face recognition in infants. (3) The hypothesis that visual illusions and aftereffects occur as a side effect of the adapting lateral interactions in the system is tested computationally. (4) The hypothesis that perceptual grouping occurs based on synchronization mediated by the lateral and feedback connections is tested, verifying that Gestalt phenomena emerge from self-organized interactions. The project lends strong computational support to the idea that the visual cortex is a continuously-adapting recurrent structure, and that perceptual phenomena such as face recognition, illusions, and grouping are its emergent effects. www.cs.utexas.edu/users/nn/pages/research/selforg.html
