Charles Frye

Reveal Contact Info

PhD Student


DeWeese Lab

Current Research

Neural networks have taken the world of machine learning by storm, establishing state-of-the-art performance on many tasks and extending the domain of problems computers can solve, from classifying images and translating text to playing Go and Doom from only visual input.

However, it has become clear that our understanding of just why these biologically-inspired models work is woefully incomplete. My research is focused on using both analytical and empirical methods to better understand neural networks, with the aim to bring these insights back to the study of the brain regions whose phenomenology inspired the original models.


Lots of folks enter the field of theoretical neuroscience from a quantitative field, like physics or computer science, and aim to bring the tools of their original discipline to bear on problems from biology and psychology. I took the road less traveled: I started out in psychology and biology, and came to theoretical neuroscience to pick up the quantitative tools I saw as necessary to answer the questions I was asking.

At the University of Chicago, I worked in the lab of Harriet de Wit on human behavioral and physiological responses to the recreational drug MDMA and in the lab of Jason MacLean on the perinatal development of neural circuit activity. These experiences continue to inform my perspective on the brain as an embodied piece of biological hardware whose principles and purposes are best understood at a computational and psychological level.

Outside of work, I write a blog, Tangent Space, that covers mainly topics in probability and statistics, with a didactic bent and a conversational style. I also have a passionate, but amateur, interest in medieval history, particularly of subaltern and interstitial groups, from heretics to itinerant shepherds. I am also a Dungeons & Dragons Dungeon Master and occasional house DJ.