Representation of visual uncertainty through neural gain variability

Hénaff, Olivier J.; Boundy-Singer, Zoe M.; Meding, Kristof; Ziemba, Corey M.; Goris, Robbe L. T.

doi:10.1038/s41467-020-15533-0

Cited by 47 publications

(101 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To control for sensory confounds and isolate specific effects of perceptual uncertainty, it should be possible to design stimulus protocols where the perceptual task is always performed with an identical probe stimulus, but where perceptual uncertainty was manipulated by prior exposure to different priming stimuli. These predictions echo recent findings that link neural gain variability to perceptual uncertainty induced by manipulations of low-level image statistics [36]. This link between uncertainty and variability is also qualitatively captured by our model (see Supplemental Fig.…”

Section: New Predictions Of Adaptive Codingsupporting

confidence: 91%

Attentional dynamics of efficient codes

Tkačik

2021

Preprint

View full text Add to dashboard Cite

Top-down attention is hypothesized to dynamically allocate limited neural resources to task-relevant computations. According to this view, sensory neurons are driven not only by stimuli but also by feedback signals from higher brain areas that adapt the sensory code to the goals of the organism and its belief about the state of the environment. Here we formalize this view by optimizing a model of population coding in the visual cortex for maximally accurate perceptual inference at minimal activity cost. The resulting optimality predictions reproduce measured properties of attentional modulation in the visual system and generate novel hypotheses about the functional role of top-down feedback, response variability, and noise correlations. Our results suggest that a range of seemingly disparate attentional phenomena can be derived from a general theory combining probabilistic inference with efficient coding in a dynamic environment.

show abstract

Section: New Predictions Of Adaptive Codingsupporting

confidence: 91%

Attentional dynamics of efficient codes

Tkačik

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Other work has established the connection between normalization and variability more directly. A descriptive model of stochastic normalization has been shown to fit changes in variability with stimulus contrast (Coen-Cagli and Solomon 2019) and orientation noise (Henaff, Boundy-Singer, et al 2020), and revealed that, even for fixed stimuli, variability is reduced during epochs of strong normalization (Coen-Cagli and Solomon 2019). Our analytical results on normalization and variability bridge the gap between this literature and a theory of the computational role of variability.…”

Section: Discussionmentioning

confidence: 99%

“…Another recent model (Henaff, Boundy-Singer, et al 2020) proposes that uncertainty is represented in the response variability, and is thus related to sampling and to our work. However, they propose that variability is partitioned into two terms, Poisson variability and fluctuations in response gain (Goris, Movshon, et al 2014).…”

Section: Discussionmentioning

confidence: 99%

“…Different from our work, the Poisson term in that framework does not have a functional role and is left unexplained, and there is no precise relation between V1 variability and the statistics of natural images. In addition, whereas sampling-based representations can approximate the full posterior distribution, the model of Henaff, Boundy-Singer, et al (2020) focuses only on the mean and variance (uncertainty) of the posterior. Therefore, future experimental work could further distinguish between these theories by comparing higher-order statistics of V1 responses to the corresponding statistics in the visual inputs.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Neuronal Variability Reflects Probabilistic Inference Tuned to Natural Image Statistics

Festa

Aschner

Davila

et al. 2020

Preprint

View full text Add to dashboard Cite

1Neuronal activity in sensory cortex fluctuates over time and across repetitions of the same input. This 2 variability is often considered detrimental to neural coding. The theory of neural sampling proposes 3 instead that variability encodes the uncertainty of perceptual inferences. In primary visual cortex (V1), 4 modulation of variability by sensory and non-sensory factors supports this view. However, it is 5 unknown whether V1 variability reflects the statistical structure of visual inputs, as would be required 6 for inferences correctly tuned to the statistics of the natural environment. Here we combine analysis of 7 image statistics and recordings in macaque V1 to show that probabilistic inference tuned to natural 8 image statistics explains Poisson-like variability, and the modulation of V1 activity and variability by 9 spatial context in images. Our results show that the properties of a basic aspect of cortical responses-10 their variability-can be explained by a probabilistic representation tuned to naturalistic inputs. 11

show abstract

“…Slow systematic changes may occur in this dataset analyzed here, but they operate at a time scale that makes them orthogonal to our theory. In particular, slow multiplicative, low-dimensional noise may serve other functional roles, such as encoding uncertainty in visual areas [26; 27], but it cannot serve as a labeling mechanism of the type proposed here. Such variability would convey information about informativeness on a time scale slower than that needed for single trial feedback and decoder learning.…”

Section: Discussionmentioning

confidence: 99%

Targeted V1 comodulation supports task-adaptive sensory decisions

Haimerl

Ruff

Cohen

et al. 2021

Preprint

View full text Add to dashboard Cite

Sensory-guided behavior requires reliable encoding of stimulus information in neural responses, and task-specific decoding through selective combination of these responses. The former has been the topic of intensive study, but the latter remains largely a mystery. We propose a framework in which shared stochastic modulation of task- informative neurons serves as a label to facilitate downstream decoding. Theoretical analysis and computational simulations demonstrate that a decoder that exploits such a signal can achieve flexible and accurate readout. Using this theoretical framework, we analyze behavioral and physiological data obtained from monkeys performing a visual orientation discrimination task. The responses of recorded V1 neurons exhibit strongly correlated modulation. This modulation is stronger in those neurons that are most informative for the behavioral task and it is substantially reduced in a control condition where recorded neurons are uninformative. We demonstrate that this modulator label can be used to improve downstream decoding within a small number of training trials, consistent with observed behavior. Finally, we find that the trial-by-trial modulatory signal estimated from V1 populations is also present in the activity of simultaneously recorded MT units, and preferentially so if they are task-informative, supporting the hypothesis that it serves as a label for the selection and decoding of relevant downstream neurons.

show abstract

Representation of visual uncertainty through neural gain variability

Cited by 47 publications

References 56 publications

Attentional dynamics of efficient codes

Attentional dynamics of efficient codes

Neuronal Variability Reflects Probabilistic Inference Tuned to Natural Image Statistics

Targeted V1 comodulation supports task-adaptive sensory decisions

Contact Info

Product

Resources

About