Associate Professor of Statistics

Sumit Mukherjee

I work at the intersection of mathematical statistics, probability, and combinatorics, with a focus on inference for dependent combinatorial data, mean-field models, random permutations, exponential random graph models, and persistence questions for Gaussian processes.

Browse publications Email Download CV Scholar profile

2014–presentColumbia Statistics faculty

2020–presentAssociate Professor

NSFDMS-2515519, DMS-2113414, DMS-1712037

Research areas

Dependent combinatorial data

Statistical theory and inference for Ising models, discrete Markov random fields, rankings, random permutations, and ERGMs.

Mean-field and Gibbs measures

Asymptotics, fluctuations, parameter estimation, large deviations, and log-concavity methods for interacting systems.

High-dimensional variational inference

Mean-field approximations and empirical Bayes methods for high-dimensional regression, GLMs, and latent variable models.

Gaussian persistence

Persistence probabilities and exponents for Gaussian processes, random polynomials, and related constrained stochastic processes.

Multilinear and tensor forms

Limit distributions for random multilinear forms, graph coloring statistics, and universality phenomena.

Network statistics

Detection, degeneracy, subgraph sampling, and inference questions for sparse graph models and network-valued data.

Recent and featured work

Fluctuations in random field Ising modelsS. Lee, N. Deb, and S. Mukherjee · to appear in Annals of Applied Probability.

Persistence and Ball Exponents for Gaussian Stationary ProcessesN. D. Feldheim, O. N. Feldheim, and S. Mukherjee · Communications on Pure and Applied Mathematics, 2025.

Gibbs Measures with Multilinear FormsS. Bhattacharya, N. Deb, and S. Mukherjee · to appear in Annals of Applied Probability.

Large random matrices with given marginsH. Lyu and S. Mukherjee · preprint.

Publications and preprints

About

2020–present

Associate Professor, Department of Statistics, Columbia University.

2014–2020

Assistant Professor, Department of Statistics, Columbia University.

2014

Ph.D. in Statistics, Stanford University; advisor: Persi Diaconis.

2007–2009

Master’s in Statistics, Indian Statistical Institute.

Sumit Mukherjee

Research areas

Dependent combinatorial data

Mean-field and Gibbs measures

High-dimensional variational inference

Gaussian persistence

Multilinear and tensor forms

Network statistics

Recent and featured work

Publications and preprints

Random permutations 5 items

Mean-field Gibbs measures 9 items

Variational inference for high-dimensional posteriors 6 items

Persistence of Gaussian processes 6 items

Limit distributions for multilinear forms 6 items

Statistics on networks 5 items

Other papers 4 items

About