Welcome!
I’m Dr. Phil Chodrow, an assistant professor in the Department of Computer Science at Middlebury College. My pronouns are he/him/his. You can email me at pchodrow@middlebury.edu
or stop by my office, Room 218 in 75 Shannon Street.
My research focuses on network science, the study of connected systems in society and nature. I draw on methods from applied mathematics, machine learning, statistics, and physics. Recent topics include models of random hypergraphs, community detection in hypergraphs, opinion dynamics on networks, and gender representation in academic science. My work is supported by the National Science Foundation.
I teach courses in math, data science, computation, and network science. I have written freely-available lecture notes for undergraduate courses on machine learning and (with Heather Zinn Brooks) network science. I have a CV. Middlebury students might want to take a look at my FAQs about classes and student research. I have thoughts about jobs at liberal arts colleges.
When I’m not working, I’m probably drinking tea, practicing aikido, playing chess, running, reading, or playing board games.
News
June 2025 |
I am an invited participant at the BIRS-CMO workshop on "Collective Social Phenomena: Dynamics and Data" in Oaxaca, Mexico. |
May 2025 |
Izabel Aguiar (SFI) and I are coorganizing a minisymposium on data, inference, and dynamics in complex social systems at the SIAM Conference on Applications of Dynamical Systems in Denver, CO. |
March 2025 |
Invited talk at Michigan State on gender representation in academic mathematics. |
February 2025 |
New preprint: "The illusion of households as entities in social networks." Izabel Aguiar (SFI) has led this project on how the concept of "households" can either clarify or distort the results of social network analysis. Very glad to have been on this intellectual journey with Izabel and senior coauthor Johan Ugander. |
In Spring '25 I am teaching CSCI 0442: Network Science and CSCI 0451: Machine Learning. | |
New preprint! I'm very pleased to share "Hyperlink Prediction via Hyperedge Copying," the fruits of two years of wonderful collaboration with Xie He (Microsoft Research) and Peter Mucha (Dartmouth). This paper poses a simple, low-dimensional mechanistic model of the formation of hypergraphs. We find that this model admits both analytic tractability and predictive performance competitive with several high-dimensional neural network approaches. It's been wonderful to take this journey together with Xie and Peter. | |
I am a nominee for Middlebury's "Inclusive Design for Learning Award," coordinated by the Center for Teaching, Learning, and Research (CTLR) and the Advisory Group on Disability Access and Inclusion (AGDAI). Congratulations to Olga Parshina, the deserving winner! | |
January 2025 |
Congratulations to student Violet Ross '25! Violet defended her thesis on growing hypergraphs with node attributes, which built on work she started with me in Summer '24 and continued through Fall '24 and Winter '25. |
New preprint! In this short note, I provide an elementary proof that the equivalence of two distinct concepts of information characterizes the class of comparison functions called Bregman divergences. This is a theoretical result that I've been curious about for a while. I'm glad to have finally figured it out! |