Radu-Alexandru Dragomir

Radu-Alexandru Dragomir is an Assistant Professor in Applied Mathematics at Télécom Paris, where he is part of the S2A research team. His academic journey includes a PhD from 2018-2021 with Jérôme Bolte and Alexandre d'Aspremont within the SIERRA team in Paris, followed by post-doctoral positions at UCLouvain with Yurii Nesterov and at EPFL in the OPTIM group of Nicolas Boumal.

Dr. Dragomir's research focuses on optimization methods for solving large-scale problems in signal processing and data science. His specific interests include:

• Non-convex optimization
• Gradient descent and its non-Euclidean variants (mirror descent)
• Dynamics of neural network training
• Ethical aspects of machine learning and artificial intelligence
• Reliability of machine learning algorithms
• Political impact of social networks and recommender systems
• Environmental cost of AI

His publication record demonstrates consistent contributions to optimization theory with recent work on convex quartic problems, implicit bias of mirror flow, and stochastic Bregman methods. His research bridges theoretical optimization with practical machine learning applications, particularly in non-Euclidean settings, with publications spanning SIAM Journal on Optimization, NeurIPS, and ICML.

Dr. Dragomir teaches several courses at Télécom Paris:

• Optimization for Machine Learning (M1)
• Signal processing tools for audio and image (L3)
• Numerical analysis (L3)
• Ecological and social transition (L3)

He currently supervises two PhD students:

• Gabriel Gros (CIFRE with Vinci Concessions, co-supervised with Olivier Fercoq)
• Yerkin Yesbay (co-supervised with Pascal Bianchi)