Abstract
The development of directional selective cortical columns requires experience and\r therefore plasticity. However, it is unknown whether this plasticity occurs within the input from\r the thalamus to the cortical columns or is a result of changing synaptic weights contained\r entirely within the cortex. Here, we examine two fixed-input cortical computational models, the\r coupled-pair and the 2-column E/I pair model, to determine if they can develop biologically\r relevant selectivities from different training paradigms, and to do so without null direction\r inhibition. We show that both models can develop direction selectivity with bidirectional and\r randomized training and do not rely on null direction inhibition. However, only the coupled-pair\r model shows biologically relevant results when trained with unidirectional training. Lastly, both\r models are consistent with cortex response sparsification during development. These results\r indicate that the coupled-pair model is the most likely candidate found within the brain.