The RF-LISSOM model has been used to examine a number of cortical phenomena. The experiments show how the observed organization of feature detectors and lateral connections in the primary visual cortex could form based on activity-dependent self-organization, driven by the regularities in the input (Miikkulainen et al. 1997; Sirosh and Miikkulainen 1995, 1996; Sirosh et al. 1996). As will be described in more detail in chapter 4, when the model is trained on retinal input patterns consisting of elongated Gaussian spots, it develops orientation columns organized into realistic orientation maps. When trained on Gaussian spots at slightly different positions on two separate receptive surfaces, the model develops realistic ocular dominance columns (areas favoring each eye). When trained on Gaussian spots of different sizes, size-selective columns develop much like those starting to be found in the cortex. In all these cases, the lateral connectivity patterns are found to follow the receptive field properties, as found in the cortex. When computational resources permit, future simulations will examine self-organization of all these parameters simultaneously. Such simulations would be trained on Gaussians varying along all of these dimensions, or on natural images processed by a model of the retina.
In addition to these developmental simulations, RF-LISSOM has been used to model cortical plasticity in the adult brain. The fundamental hypothesis is that the cortex is a continuously adapting structure in a dynamic equilibrium with both the external and intrinsic input. This equilibrium is maintained by cooperative and competitive lateral interactions within the cortex, mediated by lateral connections. As a test of this hypothesis, simulated cortical and retinal lesions were made in the model and they were shown to result in reorganization similar to that seen in monkey cortex (Miikkulainen et al., 1997; Sirosh and Miikkulainen, 1994b; Sirosh et al., 1996). This demonstrated that the self-organizing principles may still be operating in the adult brain during recovery from trauma. This thesis extends those results to show that self-organization may even influence the behavior of the intact adult brain, during normal visual processing.