Ramzi S. Dakhmouche

• Ramzi Dakhmouche, Ivan Lunati, and Hossein Gorji. Scalable Uncertainty-Aware Symbolic Regression for Network Model Discovery.
  In preparation
• Ramzi Dakhmouche, Adrien Letellier, and Hossein Gorji. Can Linear Probes Measure LLM Uncertainty?
  Conference on Neural Information Processing Systems (MLxOR), 2025.
• Ramzi Dakhmouche, Ivan Lunati, and Hossein Gorji. Network System Forecasting Despite Topology Perturbation.
  International Conference on Machine Learning (SIM), 2025.
• Ramzi Dakhmouche, Ivan Lunati, and Hossein Gorji. Noise Tolerance of Distributionally Robust Learning.
  International Conference on Machine Learning (SIM), 2025.
• Fatmazohra Rezkellah^* and Ramzi Dakhmouche^*. Adversarial Robustness via Constrained Interventions on LLMs.
  Conference on Neural Information Processing Systems (COML), 2025.
• Ramzi Dakhmouche, Ivan Lunati, and Hossein Gorji. Robust Symbolic Regression for Dynamical System Identification.
  Transactions on Machine Learning Research, 2025.
• Ramzi Dakhmouche and Hossein Gorji. Why Can't Neural Networks Master Extrapolation? Insights from Physical Laws.
  Conference on Neural Information Processing Systems (ML4PS), 2025.
• Ramzi Dakhmouche, Ivan Lunati, and Hossein Gorji. Robust Symbolic Regression for Network Trajectory Inference.
  International Conference on Learning Representations (MLGenX), 2024.

Resume

My resume can be downloaded here .

Master thesis update: September 15th, 2022

• Ericsson provisional patent filed : Multi-Task Bandits method for personalized sample efficient network parameter optimization.
• Master thesis defended.
• Machine learning paper under preparation.

Some of the classes I particularly enjoyed, this year:

As a Mathematics, Vision, Learning (MVL) Master's student at École Normale Supérieure (ENS) Paris-Saclay, these are some of the classes I particularly liked this year:

• Reinforcement learning, by Dr. Matteo Pirotta, Facebook AI research.
• Graphs in machine learning, by Dr. Daniele Calandriello, DeepMind Paris.
• Computational statistics, by Prof. Stéphanie Allasonnière, Université Paris Cité.
• Learning for time series, by Prof. Laurent Oudre , ENS Paris-Saclay.
• Advanced topics in Markov chains, by Prof. Eric Moulines, Ecole Polytechnique.
• Large scale optimization, by Dr. Emilie Chouzenoux, Inria Saclay.

Medium Articles

I have recently started writing medium articles and tutorials in which I share and explain classical but also state-of-the-art machine learning algorithms, methods and insights. Feel free follow me on medium if you're interested by such topics !

Proximal Gradient Optimization using Stochastic Approximation

Ideas behind its proof of convergence.

Topological Uncertainty for deep learning models

An interesting a posteriori confidence score.

Control of the Generalization Error in Statistical Learning theory (part 2)

Article discussing generalization inequalities for ML algorithms.

Control of the Generalization Error in Statistical Learning theory (part 1)

Article discussing generalization inequalities for ML algorithms.

Elementary analysis of a time series

Tutorial on a classical statistical prediction method, which sometimes serves as a good benchmark.

Some quotes I find interesting