Contact

The University of Texas at Austin

Email: matthewfaw [at] utexas.edu

About Me

I am a final-year Ph.D. candidate in the ECE department at The University of Texas at Austin, where I am very fortunate to be advised by Sanjay Shakkottai and Constantine Caramanis.

Prior to joining UT Austin, I spent four years at Duke University, where I graduated with majors in Electrical & Computer Engineering, Computer Science, and Mathematics. During my time there, I had the great fortune of working with Ben Lee on designing economically sustainable algorithms for datacenter-level computational sprinting, Richard Fair on designing digital microfluidic devices capable of manipulating cells within a droplet, and Nick Buchler on using CRISPR technologies to create cells which perform logical operations.

Research Interests

My interests lie broadly in the design and analysis of stochastic optimization and sequential decision-making algorithms. I am particularly interested in:

• The power and limitations of adaptive optimization algorithms (i,ii)
• Fairness in sequential decision-making, and decision-making in the presence of biased feedback
• Online algorithms (e.g., prophet inequalities) and learning with limited information and dynamic feasiblity constraints

News

• [03/04/24] I am giving a talk at Georgia Tech's ARC Colloquium Seminar Series on the Power of Adaptivity in SGD.
• [12/15/23] I attended the NeurIPS'23 workshop, Algorithmic Fairness through the Lens of Time, where Jessica and I presented our poster, On Mitigating Unconscious Bias through Bandits with Evolving Biased Feedback.
• [12/07/23] I successfully gave my Progress Review at UT Austin.
• [09/28/23] I attended Allerton, where Constantine presented our work on Beyond Uniform Smoothness: A Stopped Analysis of Adaptive SGD in the Learning and Networks III session.
• [09/06/23] I am organizing the ML Tea seminar series at UT Austin this year. We'll have weekly whiteboard talks by student speakers. Feel free to reach out to me if you'd like to give a talk on your work!
• [07/15/23] I presented our work on Beyond Uniform Smoothness: A Stopped Analysis of Adaptive SGD at the Stochastic optimization session at COLT'23. Here are the slides.
• [06/09/23] I received Dr. Brooks Carlton Fowler Endowed Presidential Graduate Fellowship in Electrical and Computer Engineering from the Cockrell School of Engineering for the 2023-2024 academic year.
• [05/14/23] One paper accepted to COLT 2023, Beyond Uniform Smoothness: A Stopped Analysis of Adaptive SGD!
• [04/21/23] I presented our work on Beyond Uniform Smoothness: A Stopped Analysis of Adaptive SGD at the IFML Workshop hosted at University of Washington.

Read more...

Publications (see )

In Submission:

• On Mitigating Unconscious Bias through Bandits with Evolving Biased Feedback
  F, Constantine Caramanis, Sanjay Shakkottai, Jessica Hoffmann
  Preliminary version: NeurIPS 2023 Workshop: Algorithmic Fairness through the Lens of Time, NeurIPS'23 Workshop, New Orleans, LA.

Published:

• Beyond Uniform Smoothness: A Stopped Analysis of Adaptive SGD
  F*, Litu Rout*, Constantine Caramanis, Sanjay Shakkottai
  Conference on Learning Theory, COLT'23, Bangalore, India.
• The Power of Adaptivity in SGD: Self-Tuning Step Sizes with Unbounded Gradients and Affine Variance
  F*, Isidoros Tziotis*, Constantine Caramanis, Aryan Mokhtari, Sanjay Shakkottai, Rachel Ward
  Conference on Learning Theory, COLT'22, London, UK.
  Check out my Twitter thread on this paper.
  Here are some slides on this result, and an 18-minute recorded talk by Isidoros and me.
• Learning To Maximize Welfare with a Reusable Resource
  F*, Orestis Papadigenopoulos*, Constantine Caramanis, Sanjay Shakkottai
  ACM SIGMETRICS / IFIP PERFORMANCE 2022, SIGMETRICS'22, Mumbai, India.
• Single Sample Prophet Inequalities via Greedy-Ordered Selection
  Constantine Caramanis, Paul Dütting, F, Federico Fusco, Philip Lazos, Stefano Leonardi, Orestis Papadigenopoulos, Emmanouil Pountourakis, Rebecca Reiffenhäuser
  ACM-SIAM Symposium on Discrete Algorithms, SODA'22, Alexandria, VA, USA (virtual).
  Supersedes and merges papers (i) and (ii)
• Mix and Match: An Optimistic Tree-Search Approach for Learning Models from Mixture Distributions
  F, Rajat Sen, Karthikeyan Shanmugam, Constantine Caramanis, Sanjay Shakkottai
  Advances in Neural Information Processing Systems, NeurIPS'20, (virtual).
• Computational Sprinting: Architecture, Dynamics, and Strategies
  Seyed Majid Zahedi, Songchun Fan, F, Elijah Cole, Benjamin C. Lee
  ACM Transactions on Computer Systems, TOCS'17, Volume 34 Issue 4, January 2017.

Reviewing

• [2023] NeurIPS, ALT, JMLR
• [2022] AISTATS (Top 10% Reviewer), ICLR (Highlighted Reviewer), NeurIPS (Top 10% Reviewer)
• [2021] AISTATS, ICML, NeurIPS

Co-authors

My coauthors include: Constantine Caramanis, Elijah Cole, Paul Dütting, Songchun Fan, Federico Fusco, Jessica Hoffmann, Philip Lazos, Benjamin Lee, Stefano Leonardi, Aryan Mokhtari, Orestis Papadigenopoulos, Emmanouil Pountourakis, Rebecca Reiffenhäuser, Litu Rout, Rajat Sen, Sanjay Shakkottai, Karthikeyan Shanmugam, Isidoros Tziotis, Rachel Ward, and Seyed Zahedi.