Behavior emerges from neural circuits, and understanding cognition requires models that respect both computational principles and biological constraints.

We develop neural network models constrained by circuit architecture (e.g., basal ganglia-ACC loops, hippocampal circuits) and resource-based regularization (e.g., rotational dynamics) to explain and predict the representational dynamics and geometry observed in EEG and fMRI—bridging behavior, neural population activity, and circuit-level mechanisms.