El Mehdi Saad

KAUST Optimal Lab E-mail: saad_mehdi_111 [at] yahoo [dot] fr

About me

Since September 2024, I have been a postdoctoral researcher at the Optimal Lab at KAUST. Prior to this, I was an Assistant Professor at the Laboratoire des Signaux et Systèmes at CentraleSupélec. Before that, I worked as a postdoctoral researcher at Inrae Montpellier on the ASCAI project, led by Nicolas Vezelen and Alexandra Carpentier. I obtained my PhD in 2022 from the Laboratoire de Mathématiques d'Orsay at Paris-Saclay University, under the supervision of Gilles Blanchard and Sylvain Arlot. My research focuses on Online Learning and Bandit Theory, with an emphasis on theoretical aspects and algorithmic development.

Research

My research interests include

• Online Learning

• Bandit Theory

• Active Learning

Publications and Preprints

• Dual Averaging Converges for Nonconvex Smooth Stochastic Optimization
  Tuo Liu, El Mehdi Saad, Wojciech Kotłowski, Francesco Orabona
  Preprint arxiv

• New Lower Bounds for Stochastic Non-Convex Optimization through Divergence Composition
  El Mehdi Saad, Weicheng-Lee, Francesco Orabona
  COLT 2025 arxiv

• ATA: Adaptive Task Allocation for Efficient Resource Management in Distributed Machine Learning
  Artavazd Maranjyan, El Mehdi Saad, Peter Richtárik, Francesco Orabona
  ICML 2025 arxiv

• On Weak Regret Analysis for Dueling Bandits
  El Mehdi Saad, Alexandra Carpentier, Tomáš Kocák, Nicolas Verzelen
  Neurips 2024 arxiv

• Covariance Adaptive Best Arm Identification
  El Mehdi Saad, Gilles Blanchard, Nicolas Verzelen
  Neurips 2023 arxiv

• Active Ranking of Experts Based on their Performances in Many Tasks
  El Mehdi Saad, Nicolas Verzelen, Alexandra Carpentier
  ICML 2023 (oral presentation) proc

• Contributions to Frugal Learning
  El Mehdi Saad
  PhD Thesis proc

• Constant regret for sequence prediction with limited advice
  El Mehdi Saad, Gilles Blanchard
  ALT 2023 proc

• Fast rates for prediction with limited expert advice
  El Mehdi Saad, Gilles Blanchard
  NeurIPS 2021    proc

• Online orthogonal matching pursuit
  El Mehdi Saad, Gilles Blanchard, Sylvain Arlot
  Preprint arxiv