Biology subjects

Pillow, J.

Publications and source records attributed to Pillow, J..

4 recordsLinked to original sources

bioRxiv2018

Sparse-Coding Variational Auto-Encoders

The sparse coding model posits that the visual system has evolved to efficiently code natural stimuli using a sparse set of features from an overcomplete dictionary. The original sparse coding model suffered from two key limitations, however: (1) computing the neural response to an image patch required minimizing a nonlinear objective function via recurrent dynamics; (2) fitting relied on approximate inference methods that ignored uncertainty. Although subsequent work has developed several methods to overcome these obstacles, we propose a novel solution inspired by the variational auto-encoder (VAE) framework. We introduce the sparse-coding variational auto-encoder (SVAE), which augments the sparse coding model with a probabilistic recognition model parametrized by a deep neural network. This recognition model provides a neurally plausible feedforward implementation for the mapping from image patches to neural activities, and enables a principled method for fitting the sparse coding model to data via maximization of the evidence lower bound (ELBO). The SVAE differs from standard VAEs in three key respects: the latent representation is overcomplete (there are more latent dimensions than image pixels), the prior is sparse or heavy-tailed instead of Gaussian, and the decoder network is a linear projection instead of a deep network. We fit the SVAE to natural image data under different assumed prior distributions, and show that it obtains higher test performance than previous fitting methods. Finally, we examine the response properties of the recognition network and show that it captures important nonlinear properties of neurons in the early visual pathway.

neuroscience↗

bioRxiv2017

Majority of choice-related variability in perceptual decisions is present in early sensory cortex

While performing challenging perceptual tasks such as detecting a barely visible target, our perceptual reports vary across presentations of identical stimuli. This perceptual variability is presumably caused by neural variability in our brains. How much of the neural variability that correlates with the perceptual variability is present in the primary visual cortex (V1), the first cortical processing stage of visual information? To address this question, we recorded neural population responses from V1 using voltage-sensitive dye imaging while monkeys performed a challenging reaction-time visual detection task. We found that V1 population responses in the period leading to the decision correspond more closely to the monkeys report than to the visual stimulus. These results, together with a simple computational model that allows one to quantify the captured choice-related variability, suggest that most of this variability is present in V1 as additive noise, and that areas downstream to V1 contain relatively little independent choice-related variability.

neuroscience↗

bioRxiv2017

Combined social and spatial coding in a descending projection from the prefrontal cortex

Social interactions are crucial to the survival and well-being of all mammals, including humans. Although the prelimbic cortex (PL, part of medial prefrontal cortex) has been implicated in social behavior, it is not clear which neurons are relevant, nor how they contribute. We found that the PL contains anatomically and molecularly distinct subpopulations of neurons that target 3 downstream regions that have been implicated in social behavior: the nucleus accumbens (NAc), the amygdala, and the ventral tegmental area. Activation of NAc-projecting PL neurons (PL-NAc), but not the other subpopulations, decreased preference for a social target, suggesting an unique contribution of this population to social behavior. To determine what information PL-NAc neurons convey, we recorded selectively from them, and found that individual neurons were active during social investigation, but only in specific spatial locations. Spatially-specific inhibition of these neurons prevented the formation of a social-spatial association at the inhibited location. In contrast, spatially nonspecific inhibition did not affect social behavior. Thus, the unexpected combination of social and spatial information within the PL-NAc population appears to support socially motivated behavior by enabling the formation of social-spatial associations.

neuroscience↗

bioRxiv2017

Systematic misperceptions of 3D motion explained by Bayesian inference

People make surprising but reliable perceptual errors. Here, we provide a unified explanation for errors in the perception of three-dimensional (3D) motion. To do so, we characterized the retinal motion signals produced by objects moving with arbitrary trajectories through arbitrary locations in 3D. Next, we developed a Bayesian model, treating 3D motion perception as optimal inference given sensory noise and the geometry of 3D viewing. The model predicts a wide array of systematic perceptual errors, that depend on stimulus distance, contrast, and eccentricity. We then used a virtual reality (VR) headset as well as a standard 3D display to test these predictions in both traditional psychophysical and more naturalistic settings. We found evidence that people make many of the predicted errors, including a lateral bias in the perception of motion trajectories, a dependency of this bias on stimulus contrast, viewing distance, and eccentricity, and a surprising tendency to misreport approaching motion as receding and vice versa. In sum, we developed a quantitative model that provides a parsimonious account for a range of systematic misperceptions of motion in naturalistic environments.

neuroscience↗