Mark S. Goldman

The paper presents a neural circuit model of reinforcement learning in songbirds that demonstrates how temporal and spatial imprecision in dopaminergic signals can be overcome through fast reward prediction updates.

The paper develops two neural circuit models for evidence accumulation through sequential neuronal activity and tests them against neural recordings from four brain regions during a decision-making task.

The paper presents a computational model explaining how analog-valued decision information is transferred between neuronal populations through dynamic gating during eye movements.

This paper introduces a working memory model using dendritic bistability to robustly maintain both stimulus location and amplitude without parameter fine-tuning.

The paper uses a larval zebrafish connectome to predict neural coding of behavior and shows statistical matches with calcium imaging data.

The paper proposes a computational model of systems consolidation, showing how synaptic dynamics in two brain regions enable memory transfer while preserving older memories.

This study uses kernel-based analysis to show context-dependent deterministic processes and consistent noise in steering behavior across species and tasks.

The paper models interactions between feedback, neural activity, and plasticity in a closed-loop motor learning task, showing how plasticity signs can change with feedback strength.

This paper proposes a phase-locking mechanism for multi-level storage in short-term memory using oscillatory activity, enabling persistent and sequential activity patterns with graded amplitudes.

The paper shows that experience modifies the timing rules of synaptic plasticity, affecting learning efficiency and temporal accuracy in mice.

This paper links oculomotor plant dynamics to neural integrator activity in larval zebrafish, showing that gaze stabilization requires integration across multiple distributed timescales.

The paper compares cortical and thalamic inputs to the nucleus accumbens in mice during a reinforcement learning task, finding that cortical inputs represent actions and choices while thalamic inputs represent cues, with cortical sequences supporting reinforcement learning.

The paper introduces seqNMF, a software toolbox for extracting and analyzing temporal sequences in high-dimensional neural data, showing its effectiveness on simulated and real neural datasets.

The paper uses calcium imaging in zebrafish to study granule neuron signaling during optokinetic behavior, revealing heterogeneous responses and functional correlations.