Community & teaching

Mentorship & Seminars

Seminars and mentorship programs I have helped organize at UC Berkeley.

Student Probability Seminar

The UC Berkeley Student Probability Seminar is a venue for graduate students in mathematics, statistics, and neighboring departments to study topics in modern probability theory. In Fall 2023, I organized the seminar with Daniel Raban around Sourav Chatterjee's Superconcentration and Related Topics.

September 6

Introductory talk — Vilas Winstein

Superconcentration, chaos, and multiple valleys across a collection of running examples.

September 13

Markov semigroups and Poincaré inequalities — Adam Jaffe

Two fundamental tools for the semester, with applications to Gaussian polymers.

September 20

Superconcentration and chaos — Vilas Winstein

The equivalence of the two phenomena in a general Markov-process framework.

September 27

Chaos implies the multiple valley property — Ella Hiesmayr

A detailed treatment for the Gaussian polymer model.

October 4

Talagrand's method — Zoe McDonald

The L¹–L² method as a sharpening of the Poincaré inequality.

October 11

Two examples of superconcentration — Mriganka Basu Roy Chowdhury

First-passage percolation and the Sherrington–Kirkpatrick model.

October 18

The spectral method for proving superconcentration — Zachary McNulty

Chapter 6: spectral decomposition of the Ornstein–Uhlenbeck semigroup, an improved Poincaré inequality, and superconcentration in the Sherrington–Kirkpatrick model.

October 25

Extremal fields — Daniel Raban

Chapter 8: superconcentration and extremality, with applications to spin glasses and the discrete Gaussian free field.

November 1

Further applications of hypercontractivity — Zachary McNulty

Chapter 9: largest-eigenvalue superconcentration, low-correlation fields, subfields, and Gaussian fields on tori and Euclidean spaces.

November 15

Dimensions of level sets — Victor Ginsburg

Chapter 12: level sets of extremal fields and their induced dimensions.

Presenter and date information is drawn from the seminar's archived Fall 2023 signup sheet and the organizers' records.

Other semesters

The 2024–2025 links are collected in Mriganka Basu Roy Chowdhury's seminar archive.

Directed Reading Program

Spring 2025 — Foundations of Data Science

I mentored a second Directed Reading Program focused on high-dimensional data analysis. We followed Blum, Hopcroft, and Kannan's Foundations of Data Science, spending roughly two to three weeks on each chapter.

Weeks 1–2

High-dimensional space — Chapter 2

Concentration and geometry in high dimensions, properties of the unit ball, high-dimensional Gaussians, random projections, and the Johnson–Lindenstrauss lemma.

Weeks 3–5

Best-fit subspaces and singular value decomposition — Chapter 3

Singular vectors and SVD, optimal low-rank approximation, the power method, principal component analysis, and applications to clustering and ranking.

Weeks 6–8

Random walks and Markov chains — Chapter 4

Stationary distributions, Markov chain Monte Carlo, Metropolis–Hastings and Gibbs sampling, conductance and convergence, electrical networks, and PageRank.

Weeks 9–11

Machine learning — Chapter 5

The perceptron and kernel methods, generalization and overfitting, regularization, online learning, support-vector machines, VC dimension, boosting, and stochastic gradient descent.

Fall 2023 — Time Series Analysis by State Space Methods

In Fall 2023, I mentored a student through Berkeley Mathematics' Directed Reading Program. We studied Durbin and Koopman's Time Series Analysis by State Space Methods, moving from the local-level model to general linear Gaussian state-space models.

September 14

Filtering and smoothing in the local-level model

Kalman filtering from conditional multivariate normal distributions, Bayesian interpretation, and state-space smoothing.

September 21

Missing observations, initialization, and parameter estimation

Forecasting with missing data, diffuse priors, and maximum-likelihood estimation of variance parameters.

September 28

The general linear model

A framework encompassing local-level, ARMA/ARIMA, exponential smoothing, seasonal, and trend models.

October 5

Initialization of the filter and smoother — Chapter 5

Exact initialization of Kalman filtering and smoothing recursions when components of the initial state are known, diffuse, or treated as unknown constants.

October 12

Maximum-likelihood estimation of parameters — Chapter 7

Likelihood evaluation through the Kalman filter, parameter estimation under diffuse initial conditions, goodness of fit, and diagnostic checking.

October 19

Illustrations of the linear model — Chapter 8

Applications of linear state-space methods to structural time series, ARMA models, spline smoothing, and dynamic factor analysis.

October 26

Special cases of nonlinear and non-Gaussian models — Chapter 9

Exponential-family and heavy-tailed models, stochastic volatility and other financial models, and broader nonlinear state-space formulations.