Vidya Muthukumar


PhD Student
Department of Electrical Engineering and Computer Sciences
University of California, Berkeley

vidya.muthukumar (at) eecs (dot) berkeley (dot) edu
264 Cory Hall

Berkeley Laboratory for Information and System Sciences (BLISS)


Excited to be starting as an Assistant Professor in the ECE and ISyE departments of Georgia Tech in January 2021!
New preprint comparing classification and regression tasks in highly overparameterized regimes.
Excited to be spending Fall 2020 at the Simons Institute program on ‘‘Theory of Reinforcement Learning“ as a Simons-Berkeley Research Fellow!
Paper on harmless interpolation in noisy linear regression to appear in IEEE Journal of Selected Areas in Information Theory, special issue on ‘‘mathematics of deep learning”.
Paper on online model selection for linear bandits to appear at Conference on Artificial Intelligence and Statistics (AISTATS) 2020.
Invited talk at Fields Institute Conference on Data Science and Optimization, November 2019.
Invited talk at Microsoft Research-New York City, November 2019.
Invited talk at INFORMS Annual Meeting, October 2019.
Invited talk at Georgia Institute of Technology ML seminar, October 2019.
Attended NextProf Nexus Workshop for future faculty at Georgia Institute of Technology.

About me

I am a final year graduate student in the EECS Department at UC Berkeley, advised by Anant Sahai. My thesis committee members are Jean Walrand, Peter Bartlett, and Shachar Kariv. I recently completed a wonderful internship at IBM Research AI as a Science for Social Good Fellow for the summer, primarily collaborating with Kush Varshney.

In January 2021, I will be joining the Electrical and Computer Engineering and Industrial and Systems Engineering departments at Georgia Institute of Technology as an assistant professor.

Research interests

My broad interests are in machine learning, game theory, mechanism design and information theory. I am interested in designing algorithms that provably adapt in strategic environments, and I additionally seek a foundational understanding of the accuracy and transparency of modern machine learning. Some other topics I have worked on include theoretical and empirical studies of incentive mechanisms for spectrum sharing, and ranking from partial pairwise comparisons.

Selected publications

  • Vidya Muthukumar, Kailas Vodrahalli, Anant Sahai: Harmless interpolation of noisy data in regression
    Shorter version at IEEE ISIT 2019, to appear in Journal for Selected Areas in Information Theory, inaugural special issue on “Deep Learning: Mathematical Foundations and Applications too Information Science”.

  • Vidya Muthukumar, Tejaswini Pedapati, Nalini Ratha, Prasanna Sattigeri, Chai-Wah Wu, Brian Kingsbury, Abhishek Kumar, Samuel Thomas, Aleksandra Mojsilovic and Kush R. Varshney: Understanding Unequal Gender Classification Accuracy From Face Images
    shorter version at IEEE CVPR Workshop on Bias Estimation in Face Analytics, submitted to IEEE Transactions on Information Forensics and Security.

Teaching and service

Recently, I co-instructed a special topics course with Anant Sahai on sequential decision-making under uncertainty in Fall 2018. I TAed and co-designed the first full-scale offering of EE16A in Fall 2015 at UC Berkeley, and served as the co-president of Women in Computer Science and Electrical Engineering (WICSE) for the academic year 2016-17.