Jonathan W. Pillow

This paper introduces a probabilistic model to analyze multi-region neural dynamics and communication, showing it outperforms existing models on mouse neural datasets.

The paper introduces a method to map deep network units to real neurons and shows that visual projection neurons in Drosophila form a distributed population code for social behavior.

The paper identifies and corrects misattribution errors in calcium imaging by developing a metric, visualization tool, and robust estimator for large datasets.

The paper investigates how much choice-related variability in perceptual decisions originates from early sensory cortex by analyzing neural responses in V1 during a visual detection task.

The paper shows that dorsomedial striatal pathways oppositely control behavior depending on task demands and internal state.

The paper shows mice switch between multiple decision-making strategies during perceptual tasks, challenging the assumption of a single strategy.

The paper introduces a new spatial covariance function for fMRI data called the brain kernel, which captures non-stationary spatial correlations and improves decoding and factor analysis.

The paper introduces NAOMi, a simulation framework for evaluating calcium imaging methods using in silico mouse cortex data, comparing segmentation algorithms and optical designs.

This paper introduces a method to analyze time-varying decision-making strategies in animals and applies it to data from mice, rats, and humans.

The paper trains a recurrent neural network to model decision-making in monkeys, revealing grid-like fixed points in network dynamics for context-dependent tasks.

This paper introduces Poisson balanced spiking networks to address the instability caused by synaptic delays in traditional balanced spiking networks, showing improved robustness and realistic neural correlations.

The paper investigates multi-dimensional dynamic encoding in prefrontal cortex during decision-making in monkeys, revealing time-evolving representations of decision variables and context.

This paper introduces SNP-GPFA, a Gaussian process factor model to separate signal and noise in neural population activity, showing noise perturbs activity orthogonal to signal.

This paper introduces a conductance-based encoding model to infer synaptic inputs from spike trains and shows it outperforms traditional models in predicting retinal ganglion cell responses.

The paper develops an unsupervised method to identify internal states from behavioral data and applies it to courtship behavior in Drosophila, revealing three latent states and their neural basis.

This study shows error-correcting attractor dynamics reduce noise effects in working memory by pulling memories toward stable representations in both humans and monkeys.

This paper identifies bias in RSA methods using fMRI data and introduces BRSA, a Bayesian approach that reduces bias and improves precision in estimating neural representational similarity.

The paper compares discrete stepping and nonlinear ramping dynamics in LIP neuron spiking responses during decision-making and finds that stepping dynamics better describe most neurons, though an extended ramping model accounts for a similar fraction.

The paper compares deep learning and traditional methods for fitting linear-nonlinear-Poisson models of neural encoding, finding that deep learning methods achieve higher accuracy and require more filters.

The paper develops a Bayesian model to explain systematic perceptual errors in 3D motion perception and validates it with VR and 3D display experiments.

The paper introduces a flexible model-based approach to analyze neural variability, showing that many neurons exhibit non-quadratic mean-variance relationships and demonstrating improved adaptive stimulus selection.

The paper identifies a subpopulation of prefrontal cortex neurons that integrate social and spatial information to support socially motivated behavior.

The paper examines how running affects visual responses in higher-order visual areas of mice, finding reduced firing but improved decoding accuracy.

The paper develops scalable Bayesian methods for estimating high-dimensional neural receptive fields, achieving fast and memory-efficient inference.

The paper investigates how cortical structures generate and retrieve event representations during continuous narratives, revealing event boundaries and their neural correlates.

The paper responds to criticisms of previous findings on LIP neuron dynamics by reanalyzing data with lower bounds and showing the stepping model still better explains spike trains in many cases.

This paper introduces vTwINS, a volumetric two-photon calcium imaging method that increases the number of neurons recorded while maintaining high frame-rate in mouse brain regions.