Skip to main content

Michael Puthawala

Picture of Michael Puthawala

Title

CAPITAL Services Scholar in Artificial Intelligence and Machine Learning Assistant Professor

Office Building

Chicoine Architecture, Mathematics and Engineering Hall

Office

248

Mailing Address

Chicoine Architecture, Math & Engineering Building 248
Math & Statistics-Box 2225
University Station
Brookings, SD 57007

Biography

Michael Puthawala is the CAPITAL Services Scholar in Artificial Intelligence and Machine Learning assistant professor. He is an applied mathematician working in the field of machine learning, with an emphasis on mathematical machine learning, especially manifold learning, topological/geometric learning and universality. He also has interests in more classical math/applied math topics, for example inverse problems, scientific computing and optimal transport.

Education

  • Ph.D. in applied mathematics | University of California, Los Angeles | 2019
  • M.S. in applied mathematics | University of California, Los Angeles | 2016
  • B.S. in mathematics | Rensselaer Polytechnic Institute | 2014

Academic Interests

  • Machine learning: manifold learning, geometric learning, universality
  • Math/applied math: inverse problems, scientific computing, optimal transport

Work Experience

  • 2019-2022 Simons Postdoctoral Fellow, Rice University Department of Computational Math and Operations Research
  • 6/2018 - 9/2018 Summer Software Research Intern, Google LLC. Venice, California
  • 5/2017 and 5/2016 Summer Research Intern, Oak Ridge National Lab. Oak Ridge, Tennessee 
  • 6/2014 - 8/2014 and 6/2013 - 8/2013 Summer Research Intern, MIT Lincoln Lab. Lexington, Massachusetts 

Areas of Research

  • Manifold Learning
  • Topological/Geometric Learning
  • Universality
  • Optimal Transport
  • Inverse Problems

Publications

  • Deep Invertible Approximation of Topologically Rich Maps between Manifolds - M. Puthawala, M. Lassas, I. Dokmanic, P. Pankka, M. de Hoop
  • Universal joint approximation of manifolds and densities by simple injective flows - M. Puthawala, M. Lassas, I. Dokmanic, M. de Hoop - International Conference on Machine Learning, 17959-17983
  • Globally injective ReLU networks - M. Puthawala, K. Kothari, M. Lassas, I. Dokmanic, M. de Hoop - Journal of Machine Learning Research 23 (105), 1-55
  • Unnormalized optimal transport - W. Gangbo, W. Li, S. Osher, M. Puthawala - Journal of Computational Physics 399, 108940

Department(s)

Related Links

Google Scholar