A connectivity-constrained computational account of topographic organization in primate high-level visual cortex
We introduce the Interactive Topographic Network (ITN), a computational framework for modeling cortical organization of high-level vision. Through simulations of ITN models, we demonstrate that the topographic clustering of domains in primate inferotemporal cortex may arise from the demands of visual recognition under biological constraints on the wiring cost and modulatory sign of neuronal connections. The learned organization of the model is highly specialized but not fully modular, capturing many of the properties of organization in higher-order primates. Our work is significant for cognitive neuroscience, by providing a domain-general developmental account of topographic functional specialization, and for computational neuroscience, by demonstrating how well-known biological details can be incorporated into neural network models to account for empirical findings. Code to reproduce our results and to develop and test new ITN models is available at <https://www.github.com/viscog-cmu/ITN>. Simulation results have been deposited in Kilthub (<https://doi.org/10.1184/R1/17131319>).