Teaching

Courses Taught at Marquette University

Fall 2023

Spring 2023

  • Math 2451: Differential Equations (syllabus)
  • Math 4650/MSSC 5650: Theory of Optimization (syllabus)

Fall 2022

  • Math 2455: Differential Equations for Biomedical and Civil Engineers (syllabus)
  • MSSC 6040: Applied Linear Algebra (syllabus)

Spring 2022

  • Math 2451: Differential Equations (syllabus)
  • Math 4931/MSSC 5931: Mathematics of Medical Imaging (syllabus)

Fall 2021

Spring 2021

  • Math 2451: Differential Equations (syllabus)
  • Math 4650/MSSC 5650: Theory of Optimization (syllabus)

Fall 2020

  • Math 2455: Differential Equations for Biomedical and Civil Engineers (syllabus)

Guest Lectures

Course: Machine Learning for Biomedical Informatics, U. Chicago Graham School, Summer and Fall 2019

Topic: Deep Learning in
 Biomedical Imaging

This three hour lecture is an introduction to deep learning as used in the modern biomedical imaging pipeline. The lecture contains three parts: (1) a brief introduction to the modern convolutional neural network "zoo", (2) a discussion of successful network architectures used to solve challenging image analysis problems in biomedical imaging, including image segmentation with the U-net, and (3) a brief introduction to the problem of image reconstruction in biomedical imaging, and the potential for deep learning methods in this area, including the use generative adversarial networks (GAN's) and unrolled optimization.

Download: (Slides)


Course: Matrix Methods for Signal Processing, Data Analysis and Machine Learning, U. Michigan EECS Department, Fall 2017

Topic: Low-rank Matrix Completion

This two hour lecture is an introduction to first-order convex optimization algorithms by way of the low-rank matrix completion (LRMC) problem. The lecture begins by motivating the LRMC problem via the "Netflix prize" and other missing data problems, then moves into various optimization formulations of LRMC, including the convex nuclear norm relaxation. Next, I show how to apply the proximal gradient descent algorithm to this problem, which results in the intuitive iterative soft-thresholding algorithm (ISTA). Finally, I illustrate how the convergence of ISTA can be dramatically improved using acceleration techniques, such as fast ISTA (FISTA) and the proximal optimal gradient method (POGM). This lecture includes an interactive Julia notebook to illustrate the differences between ISTA, FISTA, and POGM.

Download: (Slides)   (Lecture Notes)   (Julia Notebook)

Note: An updated version of the Julia notebook above has been included in recent version of this course as taught by Jeff Fessler. See course page here.