Johns Hopkins School of Medicine

The Dudman Lab studies how neural circuits implement the computations that allow animals to learn from experience and generate flexible, goal-directed behavior. We are particularly interested in how reinforcement learning principles are instantiated in real biological circuits and how circuit dynamics, synaptic plasticity, and neuromodulation give rise to adaptive control of action.

A central focus of the lab is the basal ganglia and its interaction with cortex and midbrain dopamine systems. We ask how these circuits compute decisions, regulate movement vigor, and acquire novel skills. Rather than treating reinforcement learning as an abstract algorithm, we aim to identify its physical implementation: which cell types carry specific signals, how circuit architecture shapes learning rules, and how biophysical constraints modify canonical models.

Our work tightly integrates theory and experiment. We develop computational models that span normative reinforcement learning frameworks, dynamical systems models of circuit activity, and mechanistic models of synaptic plasticity and directly test their predictions in behaving mice. Using quantitative behavioral assays, large-scale electrophysiology, optical methods, and cell-type-specific perturbations, we dissect how neural population dynamics support credit assignment, exploration–exploitation tradeoffs, and adaptive regulation of movement parameters.

For graduate students, the lab offers an environment where computation and experiment are inseparable. Projects often begin with a theoretical question — for example, how policy updates are constrained by circuit architecture, or how gain control emerges from opponent pathways — and evolve through iterative modeling and empirical testing. Students are encouraged to build and analyze models, develop quantitative behavioral tasks, and engage deeply with both systems neuroscience and machine learning theory.