Ali Ezzeddine

About

Ali Ezzeddine’s research lies at the intersection of robotics, machine learning, and human-robot interaction, with a core focus on enabling robots to learn efficiently from imperfect human guidance. His major contributions center on inverse reinforcement learning (IRL) and Bayesian policy improvement, addressing two critical challenges in programming by demonstration: sparse and non-optimal demonstrations. In his most cited work, “Learning from Demonstrations and Human Evaluative Feedbacks” (2020, 20 citations), Ezzeddine pioneered a framework that combines IRL with human evaluative feedback to handle both sparsity and imperfection in demonstrations, allowing robots to infer reward functions even from limited or noisy inputs. His earlier paper, “Combination of learning from non-optimal demonstrations and feedbacks using inverse reinforcement learning and Bayesian policy improvement” (2018, 12 citations), further advanced this approach by integrating Bayesian methods to robustly refine policies from suboptimal human data. These contributions have practical implications for assistive robotics and autonomous systems, where perfect demonstrations are rarely available. Ezzeddine’s work is notable for its pragmatic, human-centric design, making robot learning more accessible and reliable in real-world settings.

Research Focus

Key Achievements

Top Papers

1. 1
   Learning from Demonstrations and Human Evaluative Feedbacks: Handling Sparsity and Imperfection Using Inverse Reinforcement Learning Approach

   20 citations · 2020
2. 2
   Combination of learning from non-optimal demonstrations and feedbacks using inverse reinforcement learning and Bayesian policy improvement

   12 citations · 2018

Key Collaborators

Contact & Links