Jonathan W. Pillow

Letting sensory codes optimize any posterior-based loss makes information maximization (infomax) one special case; blowfly large monopolar cell responses match p = 1/2 Lp reconstruction-error codes better than infomax.

PIANO integrates single-cell data across species and modalities, models gene expression, and scales to 100 million cells for drug perturbation analysis.

This study examines age-related changes in neural and behavioural variability during a decision-making task in mice, finding increased neural variability and reduced variability quenching in specific brain regions.

This paper develops compact deep neural network models that predict macaque V4 neural responses with high accuracy while using significantly fewer parameters than existing models.

The paper develops an internal state model to analyze decision-making in mice, finding that behavior is governed by four states with distinct performance and bias characteristics.

The paper introduces PIANO, a variational autoencoder for multi-omics integration, showing improved performance on single-species and multi-atlas datasets.

The paper develops new statistical tools to analyze high-dimensional neural responses in mouse visual cortex, revealing a broken power law structure and highlighting the role of inhibitory neurons in shaping population representations.

The paper develops improved models for inferring latent neural states from calcium imaging data by integrating spiking and calcium dynamics directly into latent variable frameworks.

This study infers reinforcement learning rules from single-animal behavior during de novo task learning in mice, finding that policy-gradient-like rules outperform classical temporal-difference algorithms.

Mice infer stimulus-side priors from recent choices; prior signals are decodable pre-stimulus in 24.0% to 30.2% of 242 brain regions, including early visual and motor areas.

The paper presents a brain-wide neural activity map from 621,733 neurons across 139 mice performing a decision-making task, showing how neural activity encodes task variables across brain regions.

This paper introduces PRISM plasticity, a unified model of synaptic plasticity rules, and shows it enables rapid learning of flexible neural dynamics and solves temporal credit assignment in machine learning.

This paper introduces a continuous model to partition neural variability into stimulus-driven and stochastic components, finding that gain processes follow an exponentiated power law across visual cortical areas.

This study uses facial features from mice and monkeys to infer internal cognitive states during a foraging task, finding comparable state-task relationships across species.

The paper shows that observing aggression leads to neural and behavioral changes similar to direct experience, with shared plasticity mechanisms in aggression-related networks.

The paper uses a dynamic latent state model to analyze mice learning a visual decision-making task, finding distinct internal states emerge early in training.

This paper introduces dDIRL to model complex animal behavior over time, identifying internal states and reward functions from observed navigation in water-starved mice.

The paper shows that pre-existing visual responses in dopamine terminals predict individual learning trajectories in mice during a visual task.

This paper introduces a sparse-coding variational auto-encoder (SVAE) that improves upon sparse coding by using a probabilistic model with a deep neural network, achieving better performance on natural image data.

The paper develops a connectome-constrained dynamical model for C. elegans that captures causal neural interactions and shows anatomical connectivity is sufficient for accurate predictions.

The paper develops a framework to model neural and behavioral data together, revealing distinct evidence accumulation dynamics in three rat brain regions.

The paper develops a simulation-based model to infer neural dynamics of memory during fly courtship, finding that adaptive population mechanisms transform song history into a mnemonic format.

The paper introduces a cell-type dynamical systems model to better capture neural dynamics by considering distinct cell classes, showing improved accuracy and biological interpretability in decoding behavior and simulating perturbations.

This paper introduces a Gaussian process multi-class decoder (GPMD) for neural decoding, achieving state-of-the-art accuracy across species and demonstrating the importance of correlation structure.

This paper introduces FINDR, a deep learning method that infers low-dimensional nonlinear dynamics from neural data, outperforming existing methods in capturing neuron responses and visualizing flow fields.

The paper introduces a method to learn a causal model of the fly brain using a connectome and optogenetic perturbation data, revealing that global brain dynamics arise from small, localized circuits.

This study shows that temporal integration is a common feature in perceptual decision-making across species, using behavioral data from monkeys, rats, and humans.

The paper introduces a scalable variational inference method for estimating low-rank spatio-temporal receptive fields in neural data, showing improved performance over existing estimators.

The paper describes a modular data architecture for a neuroscience collaboration, enabling data contribution, access, and analysis across multiple labs.

This paper develops a theoretical framework explaining individual variability in decision-making by analyzing neural and behavioral data from rats performing a context-dependent task.