Jon Wittmer

My research lies at the intersection of large scale distributed computing, deep learning, optimization, and inverse problems. Deep learning has recently been (ab)used to ‘‘solve’’ problems in scientific computing that it is fundamentally not well-suited for. Significant progress has been made over the last century to develop both the theory and practice of numerical methods to address challenges in the scientific community. Rather than disregarding the foundation from which the field of scientific computing is built, I seek to augment traditional methods with modern deep learning techniques to address computational challenges facing the scientific community today. In particular, I am interested in mathematically rigorous designs of machine learning systems to solve inverse problems and, when appropriate, using machine learning to replace expensive-to-compute parts of existing algorithms. One guiding principal in my work comes from Google’s remarks on the size and quality of a dataset: ‘‘Google has had great success training simple linear regression models on large data sets.’’ With this in mind, I am interested in finding places where machine learning techniques can have the highest impact with low sensitivity to the complexity of the model. I am convinced that in many cases machine learning coupled with robust traditional numerical methods can lead to excellent results, enabling the solution to previously intractable problems.

Interests

Education

M.S. Computational Science, Engineering, and Mathematics, 2021
University of Texas at Austin
B.S. Aerospace Systems Engineering & Applied Mathematics, 2019
University of Akron