Emmarius Delar
PhD Student in Reinforcement Learning • PSL Research University (Paris Dauphine)
Research Interests
Machine Learning • Monte Carlo Tree Search • Combinatorial Optimization • Federated Learning • Gene Regulatory Networks • Intelligent Mesh Generation • Economics & Game Theory
Education
PhD in Reinforcement Learning
PSL Research University (Paris Dauphine)
Advised by Prof. Tristan Cazenave.
MSc Artificial Intelligence, Systems, Data
PSL Research University (Dauphine, ENS Ulm, Mines Paris)
Highly selective program with strong theoretical and practical foundations in AI, covering Learning Theory, Optimization, Inverse Problems, Kernel Methods, and Mathematics of Deep Learning.
MS1 Applied Mathematics and Statistics
University of Rennes
GPA: 3.7/4 (with high honours • rank 2)
Magistère Statistics and Economics Modelling
University of Rennes
Highly selective program in advanced statistical and economic modeling. GPA: 3.7/4 (with high honours)
BSc Mathematics and Computer Science
University of Rennes
Minor in Economics
Research Experience
AI Research Intern
LAMSADE, Paris & Orange Labs, Châtillon
Contributing to the TREES project (TowaRds Energy Efficient diStributed learning for 6G), led by Orange and supervised by Tristan Cazenave, Morgan Chopin, and Nancy Perrot.
Conducting research on Monte Carlo Tree algorithms to develop efficient methods for optimizing network topologies in Federated Learning environments.
AI Research Intern
CERFACS, Toulouse
Continued research on uncertainty reduction for atmospheric dispersion models, supervised by Eliott Lumet and Mélanie Rochoux.
Implemented dimensionality reduction techniques using Machine Learning and Deep Learning for statistical emulation of atmospheric dispersion models.
Selected Projects
Matrix reconstruction using Sparse PCA with convex and non-convex approaches
Mar. 2025Implemented SVM and Kernel Logistic Regression for DNA sequence classification
Mar. 2025L-BFGS Optimization
Advanced optimization techniques for large-scale machine learning problems
Jan. 2025