Modelling the neural code in large populations of correlated neurons

Abstract

Neurons respond selectively to stimuli, and thereby define a code that associates stimuli with population response patterns. Certain correlations within population responses (noise correlations) significantly impact the information content of the code, especially in large populations. Understanding the neural code thus necessitates response models that quantify the coding properties of modelled populations, while fitting large-scale neural recordings and capturing noise correlations. In this paper we propose a class of response model based on mixture models and exponential families. We show how to fit our models with expectation-maximization, and that they capture diverse variability and covariability in recordings of macaque primary visual cortex. We also show how they facilitate accurate Bayesian decoding, provide a closed-form expression for the Fisher information, and are compatible with theories of probabilistic population coding. Our framework could allow researchers to quantitatively validate the predictions of neural coding theories against both large-scale neural recordings and cognitive performance.

Modeling the Diverse Effects of Divisive Normalization on Noise Correlations

“…There are numerous descriptive models of correlated neural population activity, among the most popular of these being latent variable models [Sokoloski, Aschner, and Coen-Cagli, 2021; Lin et al, 2015; Goris, Movshon, and Simoncelli, 2014; Whiteway, Averbeck, and Butts, 2020; Whiteway et al, 2019; Archer et al, 2014; Yu et al, 2008; Ecker et al, 2014], in which correlations arise from the mapping of a small number of unobserved variables onto the high dimensional neural response space. This approach effectively partitions noise correlations into underlying causes (i.e.…”

“…Moreover, this model belongs to the class of Generalized Linear Models, which are among the most widely used encoding models for neural activity [Pillow et al, 2008;Paninski, Pillow, and Lewi, 2007]. There are numerous alternative descriptive models of correlated neural population activity, among the most popular of these being latent variable models (LVMs), in which population-wide activity arises from interactions between a small set of unobserved variables [Yu et al, 2008;Ecker et al, 2014;Whiteway and Butts, 2019;Archer et al, 2014;Whiteway et al, 2019;Sokoloski, Aschner, and Coen-Cagli, 2021]. This effectively partitions the population noise covariance into underlying causes (i.e.…”

“…For this reason, successful descriptive models of neural activity have been developed to capture noise correlations [Pillow et al, 2008;Archer et al, 2014;Gardella, Marre, and Mora, 2019;Schneidman et al, 2006;Granot-Atedgi et al, 2013;Zhao and Park, 2017;Sokoloski, Aschner, and Coen-Cagli, 2021;Josić et al, 2009]. However, none of those models considers divisive normalization [Carandini and Heeger, 2012;Heeger, 1992;Albrecht and Geisler, 1991], an operation observed in a wide range of neural systems [Louie, Khaw, and Glimcher, 2013;Ohshiro, Angelaki, and DeAngelis, 2011;Olsen, Bhandawat, and Wilson, 2010] which has also been implicated in modulating the structure of noise correlations.…”

Efficient decoding of large-scale neural population responses with Gaussian-process multiclass regression

“…Since Zohary, Shadlen, and Newsome’s landmark demonstration of correlated activity in a population of MT neurons (Zohary et al, 1994), computational neuroscience has been seeking to elucidate the role that noise correlations play in the population code (Adibi et al, 2013; Bartolo et al, 2020; Cafaro & Rieke, 2010; Ecker et al, 2011; Ecker et al, 2014; Kanitscheider et al, 2015; Kohn et al, 2016; Moreno-Bote et al, 2014; Nirenberg & Latham, 2003; Nogueira et al, 2020; Panzeri et al, Schneidman et al, 2003; Sokoloski et al, 2021). Noise correlations refer to statistical dependencies in the trial-to-trial fluctuations in population activity elicited in response to a fixed stimulus, and they may either increase or decrease information relative to a population with conditionally independent neurons (Averbeck et al, 2006; da Silveira & Rieke, 2020).…”

“…Schneidman et al, 2003; Sokoloski et al, 2021). Noise correlations refer to statistical dependencies in the trial-to-trial fluctuations in population activity elicited in response to a fixed stimulus, and they may either increase or decrease information relative to a population with conditionally independent neurons (Averbeck et al, 2006; da Silveira & Rieke, 2020).…”