The structured ‘low temperature’ phase of the retinal population code

Abstract

Recent advances in experimental techniques have allowed the simultaneous recordings of populations of hundreds of neurons, fostering a debate about the nature of the collective structure of population neural activity. Much of this debate has focused on the empirical findings of a phase transition in the parameter space of maximum entropy models describing the measured neural probability distributions, interpreting this phase transition to indicate a critical tuning of the neural code. Here, we instead focus on the possibility that this is a first-order phase transition which provides evidence that the real neural population is in a ‘structured’, collective state. We show that this collective state is robust to changes in stimulus ensemble and adaptive state. We find that the pattern of pairwise correlations between neurons has a strength that is well within the strongly correlated regime and does not require fine tuning, suggesting that this state is generic for populations of 100+ neurons. We find a clear correspondence between the emergence of a phase transition, and the emergence of attractor-like structure in the inferred energy landscape. A collective state in the neural population, in which neural activity patterns naturally form clusters, provides a consistent interpretation for our results.

Ensemble inhibition and excitation in the human cortex: An Ising-model analysis with uncertainties

“…The figures presented in this section are based on the temporal bin size ∆t = 50 ms, thus striking a good balance between capturing correlations among neurons and providing enough data for analysis. This value of ∆t is slightly larger than the conventional 20 ms window used for retinal neurons [10,21,23,58], reflecting the sparsity of activity patterns in the temporal cortex [59]. We repeated our analysis for both smaller (∆t = 20 ms) and larger (∆t = 100 ms) time bins, and confirmed that our conclusions also hold on these time scales.…”

A Biologically Plausible Mechanism to Learn Clusters of Neural Activity

“…In previous work, we found that the strength of pairwise correlations between retinal ganglion cells was not precisely tuned to produce clusters, but was instead well within the range required [15]. The robustness of the clustered state suggests that neural populations elsewhere in the brain may also have sufficient correlation to be organized into clusters.…”

“…In fact, depending on the characteristics of neural noise, redundant population codes can be optimal for encoding information [9,10]. In this context, an appealing principle for population codes is that correlations among neurons organize neural activity patterns into a discrete set of clusters, which can each be viewed as an error-robust population codeword [11][12][13][14][15].…”

“…The retinal ganglion cell (RGC) population has sufficient redundancy and/or correlation to organize neural activity into such clusters [13,15]. Specifically, the probability landscape of retinal populations resembles a "mountain", where the summit is the all-silent state, due to the sparseness of neural activity ( Fig.…”

Clustering of Neural Activity: A Design Principle for Population Codes

“…To this end, we stimulated the retina with different natural movie clips (one of leaves blowing in the breeze, another of water in a river), again finding nearly identical heat capacities (Figure 3B). We also tested artificial stimulus ensembles, like flickering checkboards or spatially uniform flicker (Tkačik et al, 2014;Ioffe and Berry, 2017), and found that in all cases, the neural population was in a low temperature state poised close to criticality. These results have been reproduced in other labs, as well (Yu et al, 2013;Mora et al, 2015;Huang and Toyoizumi, 2016;Hahn et al, 2017).…”