Michael J. Frank

This study shows that acute stress reorganizes neural coupling in rMDD patients during decision-making, predicting future depressive symptoms.

HSSM provides a Python toolbox for hierarchical Bayesian neuro-cognitive modeling, enabling fast parameter estimation and model simulation through simulation-based inference.

This paper introduces a novel task to study decision-making under outcome uncertainty and uses computational models to analyze how people balance approach and avoidance behaviors.

This study investigates how early life adversity in mice affects reward processing and behavior, finding altered dopamine dynamics and behavioral opportunism.

This paper uses model-based EEG analysis to identify distinct neurocomputational mechanisms of learning impairments in MDD, BP, and SCZ compared to controls.

This study shows that ventral striatal astrocytes influence reinforcement learning by modulating decision noisiness and reward-guided choice through calcium dynamics.

This study shows dopamine enhances both reinforcement learning and working memory, reduces effort costs, and highlights the need to account for its effects on working memory when studying reinforcement learning.

This paper develops a basal ganglia-thalamo-cortical circuit model of adaptive learning that explains behavioral data and predicts brain region roles in complex learning.

This study shows a neural network model of prefrontal cortex and basal ganglia can adapt chunking strategies to improve working memory capacity, mimicking human performance and linking it to dopaminergic signals.

The paper uses computational modeling and neural recordings to show how theta band dynamics in basal ganglia structures adjust decision boundaries under conflict and uncertainty.

The paper introduces OpAL*, a model that demonstrates the normative advantages of dopamine and striatal opponency in learning and choice, particularly in sparse reward environments.

The paper extends the HDDM toolbox to fit a broader class of decision and reinforcement learning models using likelihood-free inference and hierarchical Bayesian regression.

The paper constructs a biophysically detailed model of the STN-GPe network to investigate how theta band activity and spiking in the subthalamic nucleus arise during response conflict.

The paper introduces the ACDC model, which learns and flexibly applies spatiotemporal sequences in different rhythms, demonstrating its ability to adapt tempo and rhythm without retraining.

The paper introduces a framework for understanding interference and generalization in neural networks through factorized, geometric control of plasticity.

The paper investigates cognitive gating operations in working memory using electrophysiological data and computational modeling, finding analogous computations across input, output, and motor gating.

This study identifies wave-like spatiotemporal patterns in dopamine axons across the dorsal striatum and links them to task contexts and instrumental controllability.

This paper introduces neural networks to approximate likelihoods for generative models in cognitive neuroscience, enabling fast posterior sampling and accurate parameter recovery.

The paper shows methylphenidate enhances cognitive effort by altering the benefit-to-cost ratio of cognitive work through striatal dopamine signaling.

The paper investigates how humans adaptively generalize task structures during reinforcement learning, finding they switch between compositional and conjunctive generalization based on task statistics.

The paper shows that the P300 EEG signal's relationship with learning depends on the statistical context of surprise, with larger P300s predicting more learning in changing contexts and less in outlier contexts.

This paper introduces a hierarchical Bayesian inference framework for simultaneous model comparison and parameter estimation in group studies, showing improved accuracy and robustness.

This study shows that reward prediction errors enhance incidental memory encoding, specifically during memoranda presentation, and this effect is not consolidation-dependent.

The paper investigates how schizophrenia affects the integration of reward probability and magnitude, finding that patients rely more on stimulus-response learning than expected value-based learning.

This study proposes chunking as a strategy for optimizing visual working memory by balancing storage capacity and recall precision, showing it improves performance while reducing precision and introducing inter-item dependencies.

The paper tests whether a PID controller can model human adaptive behavior in dynamic environments, finding it better describes decision-making than traditional models.

This paper introduces a Bayesian agent that learns to generalize task structures by clustering transition and reward functions independently, showing that performance depends on task domain statistics.

This study investigates decision-making impairments in schizophrenia by analyzing reward prediction error and expected value computations using computational modeling.

The paper investigates how fronto-basal ganglia circuitry contributes to response inhibition and conflict-induced slowing using fMRI and reinforcement-learning models.

The paper decodes dynamical neural signals in humans to reveal cooperation between cognitive and habit learning systems.