Research

Here is a short list of projects I’ve been involved with in the past or am currently working on. Note that complete publication history (with detailed citations) is located in my CV. Overall, much of my work is currently directed towards solving PDE-constrained optimization problems, especially with regards to solving nonsmooth and nonconvex problems. In the past, I’ve also worked on uncertainty quantification for biological models.

• 2024 (working projects)
  1. Coordinate Descent for Total-Variation Integer Optimal Control (with Paul Manns).
  2. Theory and Continuation Strategies for Moreau-Yosida Smoothed Problems (with Alexander Hsu, Aleksandr Aravkin, Dominique Orban)
  3. Adaptive Randomized Sketching for Dynamic Nonsmooth Optimization (with Drew Kouri and Harbir Antil - full journal submission).
  4. A Stochastic Trust-Region Method for Nonsmooth Optimization (with Aurya Javeed, Katya Scheinberg, and Drew Kouri).
  5. A Multifidelity Variational Inference Approach to Sea-Ice Inversion (with Christian Glusa and John Jakeman).

• 2023
  1. Local Convergence of an Inexact Trust-Region Method for Nonsmooth Optimization [preprint/Optimization Letters]
  2. Efficient Proximal Subproblem Solvers for a Nonsmooth Trust-Region Method (submitted to Computational Optimization and Applications) [preprint]
• 2022
  1. Adaptive Randomized Sketching for Dynamic Nonsmooth Optimization; Proceedings of the International Model Analysis Conference XLI.
  2. A Proximal Trust-Region Method for Nonsmooth Optimization with Inexact Function and Gradient Evaluations.
  3. A Levenberg-Marquardt Method for Nonsmooth Regularized Least Squares (to appear in SIAM Journal on Scientific Computing).
  4. Tsunami Early Warning from Global Navigation Satellite System Data using Convolutional Neural Networks in Geophysical Review Letters.
• 2021
  1. A Proximal Quasi-Newton Trust-Region Method for Nonsmooth Regularized Optimization, with the ArXiv version here.
  2. Comparison of Machine Learning Approaches for TsunamiForecasting from Sparse Observation. The EarchArXiv version can be found here.
• 2019
  1. Basis Pursuit Denoise with Nonsmooth Constraints. The IEEE Transactions on Signal Processing version can be found here.
  2. Relaxation algorithms for matrix completion, with applications to seismic travel-time data interpolation. The Inverse Problems version can be found here.
• 2016
  1. Validation of a Mathematical Model for Green Algae (Raphidocelis subcapitata) Growth and Implications for a Coupled DynamicalSystem with Daphnia Magna.
  2. Parameter Estimation Using Unidentified Individual Data in Individual Based Models
• 2015
  1. Uncertainty quantification in modeling HIV viral mechanics.
  2. Optimal Design for Minimizing Uncertainty in Dynamic Equilibrium Systems.
  3. Correlation of Parameter Estimators for Models Admitting Multiple Parametrizations.
• 2014
  1. Uncertainty quantification for a model of HIV-1 patient response to antiretroviral therapy interruptions.
  2. The Effects of Parameterization on Inverse Problems
• 2013
  1. Mathematical Modeling of HCV Viral Kinetics

Here is an unordered bibliography of work I’ve submitted/published/reported/talked. A more complete/detailed version is located in my CV.

• 2021 2
• 2020 1
• 2018 2
• 2017 1

2021

Comparison of Machine Learning Approaches for Tsunami Forecasting from Sparse Observations

1 minute read

We present several different methods for predicting tsunami wave height from GeoClaw runs.

A Proximal Quasi-Newton Trust-Region Method for Nonsmooth Regularized Optimization

1 minute read

We present a trust-region algorithm that computes the descent direction with a proximal gradient-like step.

Back to Top ↑

2020

Moreau-Yoshida Regularization and First Order Methods with Firedrake

11 minute read

A run-through of regularization and first order methods for solving the obstacle problem in Firedrake.

Back to Top ↑

2018

Basis Pursuit Denoise with Nonsmooth Constraints

1 minute read

We tried to solve the classic basis-pursuit denoise problem where we enforced sparsity in the solution and the observed data.

Relaxation algorithms for matrix completion, with applications to seismic travel-time data interpolation

less than 1 minute read

We developed a fast, amenable low-rank interpolation technique for travel time tomography data.

Back to Top ↑

2017

CSE 546 Machine Learning Final Project - Identifying Genre by Album Cover

6 minute read

In this project, we attempted to develop an algorithm that will classify an album’s musical genre/style based on the cover art. We pursued a color analysis w...

Back to Top ↑