Research

You can find my Google Scholar page here.

Research in Distribution Estimation

Distribution estimation concerns the problem of estimating an unknown probability distribution from a finite set of samples. The empirical estimator—equivalently, the maximum likelihood estimator (MLE)—is known to be asymptotically optimal. However, alternative procedures such as add-constant estimators, the modified Good–Turing estimator, and the profile maximum likelihood (PML) estimator can demonstrate improved performance under various choices of loss functions. These estimators are natural, in the sense that they assign equal probability to symbols that appear the same number of times in the observed sample.

In our work, we characterize the unavoidable estimation error incurred by the class of natural estimators. We further develop strategies that leverage some side information to beat the performance of natural estimators.

1. Estimating Error in Natural Distribution Estimation
   H. Balasundaram, A. Thangaraj.
   Annual Allerton Conference on Communication, Control, and Computing, 2025. Conference Paper. Slides.

2. Distribution Estimation with Side Information
   H. Balasundaram, A. Thangaraj.
   Submitted to International Symposium on Information Theory, 2026. arXiv.

Research in Information Theory

1. Learning to Transmit Over Unknown Erasure Channels with Empirical Erasure Rate Feedback
   H. Balasundaram, K. Jagannathan.
   arXiv.

Research on Stable Matchings

1. Generalized Capacity Planning for the Hospital-Residents Problem
   H. Balasundaram, G. Limaye, M. Nasre, and A. Raja.
   Theoretical Computer Science, 2026. Journal Paper.

2. Stability Notions for Hospital Residents with Sizes
   H. Balasundaram, Krishnashree J. B., G. Limaye, M. Nasre.
   Foundations of Software Technology and Theoretical Computer Science 2025. Conference Paper.