Search Papers | Poster Sessions | All Posters
Poster C99 in Poster Session C - Friday, August 9, 2024, 11:15 am – 1:15 pm, Johnson Ice Rink
Local lateral connectivity is sufficient for replicating cortex-like topographical organization in deep neural networks
Xinyu Qian1 (), Amir Ozhan Dehghani1, Asa Borzabadifarahani1, Pouya Bashivan1,2; 1McGill University, 2Mila, Université de Montréal
Across the primate cortex, neurons that perform similar functions tend to be spatially grouped together. In the high-level visual cortex, this widely observed biological rule manifests itself as a modular organization of neuronal clusters, each tuned to a specific object category. The tendency towards short connections is one of the most widely accepted views of why such an organization exists in many animals' brains. Yet, how such a feat is implemented at the neural level remains unclear. Here, using artificial deep neural networks as test beds, we demonstrate that topographical organization similar to that in the primary, intermediate, and high-level human visual cortex emerges when units in these models are laterally connected and their weight parameters are tuned using top-down credit assignment. Importantly, the emergence of the modular organization without any explicit topography-inducing learning rules and learning objectives questions their necessity and suggests that local lateral connectivity alone may be sufficient for the formation of the topographic organization across the cortex.
Keywords: modular learning topographical organization representation learning neural networks