Daniel Kunin

Personal Website

Reveal Contact Info

Postdoc

Neuroscience

DeWeese Lab

Current Research

Neural network models have revolutionized artificial intelligence, yet the mathematical foundations of their success remain unclear. My research investigates the learning dynamics of neural networks to understand how inductive biases emerge through training and how networks extract meaningful representations from data. Integrating insights from statistics, physics, and neuroscience, I aim to uncover fundamental mathematical principles governing learning in both artificial and natural intelligence.

Background

I’m a Miller Postdoctoral Fellow hosted in the Department of Neuroscience. I completed my PhD in Computational and Mathematical Engineering at Stanford University, where I was advised by Surya Ganguli. Early in my PhD, I took a seminar on the theoretical foundations of deep learning and was struck by how little was understood. As I dug deeper, I began to see artificial intelligence as a lens for studying natural intelligence, a perspective that eventually led me to the Redwood Center. Outside of research, you can find me rock climbing in the Sierras or on a whitewater canoe trip in the Canadian wilderness.