Scaling principles of distributed circuits

Abstract

Distributed circuits like the olfactory cortex, hippocampus, and cerebellum contain sub-circuits whose inputs distribute their axons over the entire circuit creating a puzzle of how information is encoded. One method for approaching the puzzle is to view them as scalable systems. In scalable systems the quantitative relationship between circuit components is conserved across brain sizes, and by mapping circuit size to functional abilities -e.g. visual acuity in the visual circuit -scientists have explained information encoding. This approach has not been applied to anti-map circuits as their scalability is unknown. To address this gap in knowledge, we obtained quantitative descriptions of the olfactory bulb and piriform cortex in six mammals using stereology techniques and light microscopy. We found that the olfactory circuit is scalable as it satises three requirements of scalable systems. First, quantitative relationships between circuit components are conserved: the number piriform neurons n scales with bulb glomeruli g as n ∼ g 3/2 . Second, the olfactory circuit has an invariant property: the average number of synapses between a bulb glomerulus and piriform neuron is one. Third, the olfactory circuit is symmorphic, i.e. olfactory ability improves with circuit size. Other distributed circuits with similar properties might also be scalable.

Connectivity and dynamics in the olfactory bulb

“…1C). This prediction is consistent with previous experimental studies [3,4], but differs from many models, which, for simplicity, treat all MCs associated to a glomerulus as equivalent [15,18,21,84,[110][111][112][113]. Thus, our results suggest the importance of accurately including network structure for determining the function of individual cells in the OB and their collective behavior, as well as the power of our model in providing a way to probe this relationship.…”

The unreasonable effectiveness of deep learning in artificial intelligence

“…Interestingly, there are many fewer long-range connections than local connections, which form the white matter of the cortex, but its volume scales as the 5/4 power of the gray matter volume and becomes larger than the volume of the gray matter in large brains (19). Scaling laws for brain structures can provide insights into important computational principles (20). Cortical architecture including cell types and their connectivity is similar throughout the cortex, with specialized regions for different cognitive systems.…”

Wiring logic of the early rodent olfactory system revealed by high-throughput sequencing of single neuron projections

“…Furthermore, identifying structure in the connectivity of the olfactory system has been impeded by the scarcity of single cell connectivity data available. While several studies addressed this challenge in insects 33, 52, 54–59 , comparable understanding of the statistics of ensembles of single olfactory bulb or piriform cortex output neuron projections in mammals is lacking 60 , despite recent advances in optical imaging-based single neuron tracing 61, 62 . To date, owing to the extensive intertwining of neural processes, axonal tracing studies reconstructed only dozens of individual bulb output neuron projections 13–16 , a number insufficient for analyzing the statistics of their branching patterns across multiple target regions.…”