Maneesh Sahani

This paper proposes a mechanistic theory of planning in prefrontal cortex using spacetime attractors, showing that recurrent neural networks trained on planning tasks recapitulate this structure.

The paper introduces WHOLISTIC, a method to image cellular activity across all organs of a vertebrate, revealing body-wide circuits and responses to stimuli.

The paper infers context-dependent neural computation mechanisms in the prefrontal cortex using linear dynamical models fitted to monkey behavioral data, revealing two distinct input selection and integration mechanisms.

The paper introduces an infinite GPFA model to infer dynamically changing latent factors in neural data, showing it captures meaningful dynamics in hippocampal place cells during spatial tasks.

The paper shows that neuronal sensitivity to acoustic context in auditory cortex of awake mice is stable over time and comparable to spectrotemporal tuning stability.

The paper analyzes residual dynamics in macaque prefrontal cortex during a saccade-based decision task to uncover recurrent neural contributions and their stability.

Optogenetic and electrical perturbations in primate motor cortex during reaching show that movement-related neural activity is governed by a self-contained, low-dimensional dynamical subspace causally linked to reach kinematics.

The paper shows that learning realigns cortical network dynamics to better integrate relevant sensory input in mice, explained by changes in excitatory-inhibitory connectivity.

The paper adds different types of recurrence to a neural network and finds that bio-inspired recurrence improves classification of noisy images, while task-trained recurrence behaves differently.

This paper demonstrates an optical brain-computer interface (oBCI) using dendritic calcium signals from rhesus macaques during motor tasks, achieving stable decoding of movement direction.

The paper investigates how sustained activation of parvalbumin-positive interneurons in auditory cortex affects divisive gain control across different stimulus types.

The paper investigates how movement initiation and grasp representation are encoded in premotor and primary motor cortex neurons during execution, observation, and withholding of reach-to-grasp actions in macaques.

The paper proposes a neural framework for online probabilistic inference in dynamic environments, demonstrating postdictive effects through distributed distributional coding.

This paper develops an ultralight head-mounted camera system for monitoring behavior and electrophysiology in freely moving mice, showing correlations between eye position, whisking, and neural activity.

The paper shows that model selection in neural data can be unreliable due to brittleness, where small data changes affect model rankings, and suggests ways to address this issue.

This paper introduces a scalable method for analyzing neural population spike trains by extending Gaussian Process Factor Analysis to handle continuous time data and capture trial-to-trial variability.

The paper demonstrates how recurrent neural networks can generate motion-selective receptive fields through temporal dynamics and learned connectivity from natural stimuli.

This study develops a deterministic spiking network model to analyze multi-neuron recordings and shows that feedback inhibition controls noise correlations in cortical networks.