Dhruva V. Raman

Small circuit principles for learning in large worlds

This paper introduces Hetlearn, a learning framework inspired by Drosophila neural circuits, which improves learning in uncertain environments with limited data and cognitive capacity.

How cerebellar architecture facilitates rapid online learning

This paper uses mathematical analysis and simulations to show that cerebellar architecture, specifically size expansions, helps overcome challenges in online learning of dynamic tasks by enabling dense granule cell activity.

Optimal synaptic dynamics for memory maintenance in the presence of noise

The paper shows that optimal memory maintenance requires systematic synaptic plasticity to be less than internal noise, predicting 50% or more of synaptic turnover is random.

Stable task information from an unstable neural population

The paper shows that neural population drift in sensorimotor cortex is systematically constrained, allowing for stable task information despite apparent instability.

Fundamental bounds on learning performance in neural circuits

The paper shows that neural circuits have an optimal size for learning performance, where larger circuits can either enhance complex behaviors or improve learning of simpler tasks, but beyond a point, increased size impairs learning due to noise.