Joel Zylberberg

Increased activity without changes in coupling between somata and apical dendrites during visual mismatches

This study examines neural responses in mouse visual cortex to natural and artificial visual stimuli, finding increased activity in response to natural mismatches but no change in dendritic-somatic coupling.

Designing stimuli for non-invasive, cell-type-specific manipulation of retinal ganglion cell activity

The paper introduces a computational framework to design stimuli that selectively manipulate specific retinal ganglion cell types with minimal impact on others.

Should I stay or should I go? Generalized marginal value theorem with temporal discounting

This paper derives an optimal patch foraging policy incorporating temporal discounting and shows how it differs from the Marginal Value Theorem under varying conditions.

Biophysical neural adaptation mechanisms enable artificial neural networks to capture dynamic retinal computation

This paper develops a deep learning model of the retina with photoreceptor adaptation mechanisms, improving prediction of primate and rat retinal ganglion cell responses to dynamic stimuli.

Stimulus-dependent functional network topology in mouse visual cortex

The paper examines how functional connectivity in mouse visual cortex changes with different stimuli, finding that motif frequencies are preserved while neuron identities within motifs shift, and functional modules become more regionally contained with complex stimuli.

Learning from unexpected events in the neocortical microcircuit

The paper shows that cortical pyramidal neurons exhibit distinct unexpected event signals in somata and distal apical dendrites that change over days, suggesting a role in hierarchical computation.

Ignoring correlated activity causes a failure of retinal population codes under moonlight conditions

The paper shows that ignoring correlated noise in retinal ganglion cells leads to poor decoding performance under moonlight conditions.

Global motion processing by populations of direction-selective retinal ganglion cells

The paper studies how populations of direction-selective retinal ganglion cells encode and decode global motion, finding that larger populations achieve high temporal precision despite sparse and broadly tuned responses.

Models of the primate ventral stream that categorize and visualize images

The paper trains deep neural networks with different objectives to model primate ventral stream computations and finds the semi-supervised model best matches neural data.

Transformation of population code from dLGN to V1 facilitates linear decoding

The paper investigates how neural representations transform from dLGN to V1 in mice, finding evidence for both sparse coding and linear separability.

A Moment-Based Maximum Entropy Model for Fitting Higher-Order Interactions in Neural Data

This paper introduces a maximum entropy model that captures higher-order neural interactions by adaptively identifying estimable moments, improving prediction of spiking pattern frequencies.

A self-organizing short-term dynamical memory network

This paper develops a self-organizing neural network with synaptic plasticity rules that enable memory formation and retention of information over extended periods.

The Role of Untuned Neurons In Sensory Information Coding

Theoretical analysis and mouse visual cortex recordings show that completely untuned neurons improve population sensory coding via inter-neuronal correlations, sometimes outperforming populations composed exclusively of tuned neurons.