Revealing nonlinear neural decoding by analyzing choices

Abstract

Sensory data about most natural task-relevant variables are entangled with task-irrelevant nuisance variables. The neurons that encode these relevant signals typically constitute a nonlinear population code. Here we present a theoretical framework for quantifying how the brain uses or decodes its nonlinear information. Our theory obeys fundamental mathematical limitations on information content inherited from the sensory periphery, describing redundant codes when there are many more cortical neurons than primary sensory neurons. The theory predicts that if the brain uses its nonlinear population codes optimally, then more informative patterns should be more correlated with choices. More specifically, the theory predicts a simple, easily computed quantitative relationship between fluctuating neural activity and behavioral choices that reveals the decoding efficiency. This relationship holds for optimal feedforward networks of modest complexity, when experiments are performed under natural nuisance variation. We analyze recordings from primary visual cortex of monkeys discriminating the distribution from which oriented stimuli were drawn, and find these data are consistent with the hypothesis of near-optimal nonlinear decoding.

Revealing unexpected complex encoding but simple decoding mechanisms in motor cortex via separating behaviorally relevant neural signals

“…Unlike their works, we focus on studying the decoding mechanism after removing the noise (irrelevant signals). Although our research perspectives differ, our results support their idea that the brain needs nonlinear operations to suppress noise interference (Yang et al, 2021)…”

“…With regard to studying decoding mechanisms, recent studies ( Pitkow et al, 2015; Yang et al, 2021 ) have focused on investigating how the brain decodes task information in the presence of noise. Unlike their works, we focus on studying the decoding mechanism after removing the noise (irrelevant signals).…”

“…And what’s the role of these ignored signals? In terms of neural decoding, irrelevant signals would significantly complicate the information readout ( Pitkow et al, 2015; Yang et al, 2021 ), potentially hindering the discovery of the true readout mechanism of behaviorally-relevant responses. Specifically, in motor cortex, in what form (linear or nonlinear) downstream neurons read out behavioral information from neural responses is an open question.…”

A distributed and efficient population code of mixed selectivity neurons for flexible navigation decisions

“…However, the brain possesses mechanisms that could enable nonlinear decoding [83][84][85] . With a nonlinear decoder, a downstream network could read out the XOR identity by combining cells with pure selectivity, that is by nonlinearly combining cells with pure sample cue selectivity and cells with pure test cue selectivity (Fig.…”

An Attempt to Analyze a Human Nervous System Algorithm for Sensing Earthquake Precursors

“…As we all know, in the background of today's supercomputers so advanced, scientists are difficult to calculate the optimal solution of nonlinear isolated wave partial differential equations. As for the human brain, if the bruteforce algorithm, fullforce algorithm, and genetic algorithm, which imitate animal neural networks, are originally special algorithms for the function of human neuronal systems, they possess more advantages than those built on linear thinking machines [29]. Then when the human brain processes entropy waves far slower than the speed of light (LFSE), it possesses more advantages than those built on linear thinking machine possesses more advantages than algorithms built on a linear thinking machine.…”