Projects
Research projects exploring neural dynamics, spiking network models, and computational neuroscience.
Decision Making in Low-Rank RNNs
Training low-rank RNNs on decision-making and working-memory tasks, building on Dubreuil et al. (2022). The connectivity statistics alone are enough to rebuild the decision network, but the working-memory network found a shortcut: it learned to keep time instead of remembering.
Slow Dynamics in Clustered Spiking Networks
From-scratch reproduction of Litwin-Kumar & Doiron (2012). A 5,000-neuron spiking network where clustered excitatory connections produce metastable dynamics and the kind of slow variability you actually see in cortex.