Nonlinear convergence boosts information coding in circuits with parallel outputs

Abstract

Neural circuits are structured with layers of converging and diverging connectivity and selectivity-inducing nonlinearities at neurons and synapses. These components have the potential to hamper an accurate encoding of the circuit inputs. Past computational studies have optimized the nonlinearities of single neurons, or connection weights in networks, to maximize encoded information, but have not grappled with the simultaneous impact of convergent circuit structure and nonlinear response functions for efficient coding. Our approach is to compare model circuits with different combinations of convergence, divergence, and nonlinear neurons to discover how interactions between these components affect coding efficiency. We find that a convergent circuit with divergent parallel pathways can encode more information with nonlinear subunits than with linear subunits, despite the compressive loss induced by the convergence and the nonlinearities when considered separately.

Retinal Encoding of Natural Scenes

“…The appropriate model structure is important not only for parameter fitting under the coding or circuit perspective but also for defining the space of potential operations in which to search for optimal designs under the normative perspective. For example, rather than modeling a single nonlinearity, normative approaches may profit from modeling retinal nonlinearities at multiple stages, which can confer an information-boosting effect (Gutierrez et al 2021).…”

Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits

“…It has been shown that BTA, or part of such architecture, is useful for efficient information 2 processing, including retaining more information (23), enhancing computational speed and accuracy (24), and amplifying variability of incoming stimuli while enhancing representation (25).…”

Temporal resolution of spike coding in feedforward networks with signal convergence and divergence

“…If a small population synapses with a much larger population, we call this structure "divergent." Examples where both structural convergence and divergence are present include the mammalian early visual system [14][15][16][17], mammalian cerebellum-like structures [18,19] and the insect olfactory system [20]. A notable example is the divergence from 200 million mossy fibers to 50 billion granule cells and then convergence to 15 million Purkinje cells in the human cerebellum, a largely feedforward network.…”