Fundamental bounds on the fidelity of sensory cortical coding

Rumyantsev, Oleg; Lecoq, Jérôme; Hernandez, Oscar; Zhang, Yanping; Savall, Joan; Chrapkiewicz, Radosław; Li, Jane; Zeng, Hongkui; Ganguli, Surya; Schnitzer, Mark J.

doi:10.1038/s41586-020-2130-2

Cited by 184 publications

(260 citation statements)

References 55 publications

Supporting

Mentioning

244

Contrasting

Order By: Relevance

“…As long as this baseline activity is low, an incoming stimulus would deactivate the RSN and shift the focus from the internal state to the external environment (Deco et al, 2013). A recent study (Rumyantsev et al, 2020) demonstrated that a signal correlated over large cortical areas and orthogonal to sensory input degraded stimulus acuity, suggesting that in a high-baseline state, the internal activity can compete with the orthogonal externally evoked signal and reduce detection.…”

Section: Discussionmentioning

confidence: 99%

Spontaneous activity competes externally evoked responses in sensory cortex

Karvat

Alyahyay

Diester

2020

Preprint

View full text Add to dashboard Cite

The activity state of neuronal populations is key to understanding information processing in the brain. Most of our knowledge about sensory processing and perception comes from studying local changes in firing rates and local field potentials (LFP) evoked by external stimuli. However, most of the cortical activity is generated intrinsically, can originate spontaneously in almost any cortical area, and spreads globally. The relationship between spontaneous activity and perception is largely unclear. Here we show that high levels of spontaneous activity in the beta-band (15-30 Hz) predict reduced detection, and that this influence can be overridden by stimulus-intensity adjustment. We found that vibrotactile stimulation of the forepaw of behaving rats evokes a robust modulation of event related field potentials and firing rates, reflected in power in low (3-10 Hz) and high (95-120 Hz) frequencies of the LFP, independent of detection. Only beta shows higher power prior and during undetected stimuli, and anti-correlates with firing rate. LFP oscillations in 15-120 Hz appear as short-high-power bursts, and by applying burst-rate detection algorithm in real-time, we found that changing the vibration amplitude according to beta-burst-rate can adjust beta's impact on detection bi-directionally. This result is supported by the finding that while in all bands bursts indicate transient synchronization of cell assemblies, only beta bursts are followed by a reduction in firing rate. Our results reveal that beta-bursts reflect increased spontaneous synchrony and predict reduced detection, suggesting a dynamic state that competes with the detection of external stimuli.

show abstract

Section: Discussionmentioning

confidence: 99%

Spontaneous activity competes externally evoked responses in sensory cortex

Karvat

Alyahyay

Diester

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Overlap of these subspaces implies that trial-to-trial variability in the noise components can corrupt the population response along the signal dimensions, thereby interfering with representation of the signal. Previous work [15] suggests that noise only interferes with the signal reprsentation if it lies in a direction defined by the derivatives of neural tuning curves, and recent work suggests that noise and signal subspaces may indeed be nearly orthogonal [14,27,28]. Because our model contains separate latent components for signal and noise, we can explicitly compare the relative angle between these subspaces.…”

Section: Visualizing Signal and Noise Subspacesmentioning

confidence: 99%

Identifying signal and noise structure in neural population activity with Gaussian process factor models

Keeley

Aoi

et al. 2020

Preprint

View full text Add to dashboard Cite

Neural datasets often contain measurements of neural activity across multiple trials of a repeated stimulus or behavior. An important problem in the analysis of such datasets is to characterize systematic aspects of neural activity that carry information about the repeated stimulus or behavior of interest, which can be considered “signal”, and to separate them from the trial-to-trial fluctuations in activity that are not time-locked to the stimulus, which for purposes of such analyses can be considered “noise”. Gaussian Process factor models provide a powerful tool for identifying shared structure in high-dimensional neural data. However, they have not yet been adapted to the problem of characterizing signal and noise in multi-trial datasets. Here we address this shortcoming by proposing “signal-noise” Poisson-spiking Gaussian Process Factor Analysis (SNP-GPFA), a flexible latent variable model that resolves signal and noise latent structure in neural population spiking activity. To learn the parameters of our model, we introduce a Fourier-domain black box variational inference method that quickly identifies smooth latent structure. The resulting model reliably uncovers latent signal and trial-to-trial noise-related fluctuations in large-scale recordings. We use this model to show that predominantly, noise fluctuations perturb neural activity within a subspace orthogonal to signal activity, suggesting that trial-by-trial noise does not interfere with signal representations. Finally, we extend the model to capture statistical dependencies across brain regions in multi-region data. We show that in mouse visual cortex, models with shared noise across brain regions out-perform models with independent per-region noise.

show abstract

“…The impact of across-neuron and across-time correlations has been long debated. Much experimental and theoretical work has proposed that correlations limit the information capacity of a neural population 6–9 . Because these correlations reflect trial-to-trial variability that is shared across neurons or time, the detrimental effect of variability on stimulus decoding cannot be eliminated by averaging the activity of neurons or time points.…”

Section: Introductionmentioning

confidence: 99%

Correlations enhance the behavioral readout of neural population activity in association cortex

Valente

Pica

Runyan

et al. 2020

Preprint

View full text Add to dashboard Cite

These correlations can take the form of correlations between 47 the spike rates of two individual cells, which we call across-neuron noise correlations. Similarly, 48 neural population activity at a given time in response to a stimulus is often correlated with the activity 49 of the same population at other times, which we refer to as across-time noise correlations. 50 51The impact of across-neuron and across-time correlations has been long debated. Much experimental 52and theoretical work has proposed that correlations limit the information capacity of a neural 53 population 6-9 . Because these correlations reflect trial-to-trial variability that is shared across neurons 54 or time, the detrimental effect of variability on stimulus decoding cannot be eliminated by averaging 55 the activity of neurons or time points. Noise correlations are therefore proposed to fundamentally 56 constrain what neural networks can compute, and limit the accuracy by which subjects can judge 57 differences between stimuli 6,7 . A reason to suspect that the effect of noise correlations may be more 58 nuanced, however, comes from a separate stream of biophysical and theoretical studies. This line of 59 work has shown that spatially and temporally correlated spiking can more strongly and more reliably 60propagate signals and drive responses in postsynaptic neural populations 10-14 . However, it remains 61 poorly tested if and how enhanced signal propagation may have a beneficial impact on behavioral 62 discrimination performance. 63 64We reasoned that we could investigate how noise correlations shape behavioral performance in 65 perceptual discrimination by studying at the same time not only how correlations impact the encoding 66 of sensory information, as has been emphasized frequently, but critically also how they impact the 67 reading out of this information by downstream neural circuits to guide behavioral choices. 68 69Correlations of PPC activity limit sensory coding during perceptual discrimination 70To examine how noise correlations affect both stimulus coding at the population level and behavioral 71 discrimination performance, we focused on the mouse posterior parietal cortex (PPC). PPC is thought 72 to participate in transforming multisensory signals into behavioral outputs and is thus likely at the 73 interface of encoding and reading out stimulus information. Furthermore, PPC is essential for 74 perceptual discrimination tasks during virtual-navigation 15,16 , and its activity has been shown to 75 contain stimulus information that relates to an animal's choices [16][17][18][19][20][21][22] . It is thus a relevant area to study 76 the impact of correlated neural activity on behavior. 77 78We examined across-time and across-neuron correlations in PPC population activity using previously 79 published calcium imaging datasets. To study across-time correlations, we used calcium imaging data 80 from a sound localization task 17 in which mice reported perceptual decisions about the location of an 81 auditory stimulus by navig...

show abstract

Fundamental bounds on the fidelity of sensory cortical coding

Cited by 184 publications

References 55 publications

Spontaneous activity competes externally evoked responses in sensory cortex

Spontaneous activity competes externally evoked responses in sensory cortex

Identifying signal and noise structure in neural population activity with Gaussian process factor models

Correlations enhance the behavioral readout of neural population activity in association cortex

Contact Info

Product

Resources

About