MRI — Mohammad Rafiqul Islam, home MRI — Mohammad Rafiqul Islam, home
  • Home
  • Research
  • Teaching
  • About
  • CV
  • Notebook

Mohammad Rafiqul Islam

Postdoctoral Research Associate · Iowa State University

I design and analyze Langevin algorithms: Markov chain Monte Carlo methods for sampling high-dimensional distributions across networks of agents, inside constrained domains, and at higher orders of accuracy. At Iowa State I work with David P. Herzog. I earned my Ph.D. at Florida State University in 2026, advised by Lingjiong Zhu.

Research & papers CV (PDF) Email

Carver Hall 461, Ames, IA Google Scholar GitHub

Portrait of Mohammad Rafiqul Islam

Research Spotlight

All research

I.current · 2026

Constrained dynamics with singular repulsion

Collision-free stochastic particle systems inside a constrained domain, and numerical schemes that stay stable near the singularity without creating artificial collisions. Joint work with D. P. Herzog.

Project overview
II.2024 – 2026

Decentralized Bayesian sampling

Agents on a network sample a shared posterior without pooling their data. This line of work removes network bias with EXTRA-type corrections and handles convex constraints through proximal regularization.

Generalized EXTRA SGLDDecentralized Proximal SGLD
III.2025

High-order Langevin Monte Carlo

Samplers built from \(P\)-th order Langevin dynamics through splitting and accurate integration. Their Wasserstein guarantees depend better on the dimension and the accuracy level as the order grows.

High-Order Langevin Monte Carlo

News

  • sep 2026Talk on scalable sampling via Langevin dynamics at the Analysis, Geometry & Probability Seminar, Iowa State.
  • aug 2026Joined Iowa State University as a postdoctoral research associate in probability.
  • jul 2026Completed my Ph.D. at Florida State: Advances in Langevin Algorithms for Decentralized, Constrained, and Accelerated Sampling.
  • apr 2026Received FSU’s Distinguished Teaching Assistant Award and the Tam Family Excellence in Graduate Student Research Award.

Selected papers

  1. Decentralized Proximal Stochastic Gradient Langevin DynamicsM. R. Islam, L. Zhu · arXiv 2026
  2. High-Order Langevin Monte Carlo AlgorithmsT. L. Dang, M. Gürbüzbalaban, M. R. Islam, N. Yao, L. Zhu · arXiv 2025
  3. Generalized EXTRA Stochastic Gradient Langevin DynamicsM. Gürbüzbalaban, M. R. Islam, X. Wang, L. Zhu · arXiv 2024
  4. GJR-GARCH Volatility Modeling under NIG and ANN for Predicting Top CryptocurrenciesF. Mostafa, P. Saha, M. R. Islam, N. Nguyen · J. Risk Financial Manag. 2021

Full publication list

Rafiq Islam teaching calculus at a chalkboard

Teaching

You played a large part in convincing me that I am capable of learning difficult subjects.
Student evaluation, MAP 2302 Ordinary Differential Equations, Summer 2026

I have taught calculus, differential equations, precalculus, and actuarial mathematics since 2018. This fall I lead Calculus III recitations at Iowa State.

Courses & student feedback

MRI Copyright 2026 @ Dr. Rafiq Islam All rights reserved License Powered by Quarto

Contact Department of Mathematics Iowa State University Room 461, Carver Hall 411 Morrill Road Ames, IA 50011 mri@iastate.edu

Connect

Email GitHub LinkedIn Google Scholar YouTube Twitter