About

I am a postdoctoral researcher at L2S, CentraleSupélec (CNRS / Université Paris-Saclay), working on statistical and computational foundations of modern machine learning. My research sits at the intersection of statistical learning theory, optimization, and computer vision, with recent work on generative modeling, domain generalization, and vision-language models for navigation.

I completed my PhD in Computer Science in 2024 at the Institute for Informatics and Automation Problems of NAS RA, advised by Arnak Dalalyan. My thesis, Statistical and Computational Complexity of the Feature Matching Map Detection Problem, studied minimax-optimal estimators for recovering correspondences between noisy point sets in the presence of outliers. I have been a researcher at YerevaNN since 2016, with prior internships at Google Research (Zurich), Google NYC, and X — the Moonshot Factory.

My work has appeared at ICML, AISTATS, CVPR, Electronic Journal of Statistics, and others. My research interests include the statistical theory of estimation, robust and invariant learning, and adapting large multimodal models to grounded tasks such as aerial navigation.

Selected publications

• Vardanyan, E., Minasyan, A., Hunanyan, S., Galstyan, T., Dalalyan, A. Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution. ICML, 2024.
• Minasyan, A., Galstyan, T., Hunanyan, S., Dalalyan, A. Matching Map Recovery with an Unknown Number of Outliers. AISTATS, 2023.
• Galstyan, T., Minasyan, A., Dalalyan, A. Optimal detection of the feature matching map in presence of noise and outliers. Electronic Journal of Statistics, 2022.
• Galstyan, T., Harutyunyan, H., Khachatrian, H., Ver Steeg, G., Galstyan, A. Failure Modes of Domain Generalization Algorithms. CVPR, 2022.

See the Publications page for a full list, or Google Scholar.