Sagar Malhotra

He said, “Leibniz was wrong. It’s obvious, everything that happens affects everything else. This is part of science.” And he said, “Furthermore, Leibniz was wrong about everything.” Then he paused and said, “But it’s just as hard to be wrong about everything as to be right about everything.”
— from an anecdote about Gödel in Journey to the Edge of Reason

I am a postdoc at the Technical University of Vienna in the Machine Learning Research Unit hosted by Prof. Thomas Gärtner. My research interests are as follows:

• Machine learning on relational data. I am interested in efficient and statistically sound machine learning algorithms for large-scale relational data like social and epidemiological networks. This work has led me to also study aspects of random graph models and enumerative combinatorics.

• Explainability and Interpretability. I believe that any notion of trustworthy explanation must be provably correct — I develop such notions of explanation. Alongside, I also develop methods that can (hopefully efficiently) audit the correctness of an explanation, or at least provide approximate guarantees.

• Efficient Statistical Verification. I am also interested in devising efficient methods for statistical verification of machine learning models against safety properties like robustness and fairness. In this regard, I focus on leveraging PAC-style guarantees from learning theory and computational geometry literature.

Previously, I did my PhD at the University of Trento and Fondazione Bruno Kessler in Italy, supervised by Luciano Serafini .