Abbas is a research scientist at DeepMind. Before DeepMind he obtained a Ph.D. in computer science in Portugal under the supervision of Luis Paulo Reis, Nuno Lau and Gerhard Neumann. Abbas has actively participated in Robocup competitions with several achievements including two world championships. His scientific interests are generally at the conjunction of machine learning, decision-making and robotics. Research topics that he is currently working on include (offline) reinforcement learning, multi-objective RL and constraint RL with application in robotics in mind.
Talk title: Multi-Objective Reinforcement Learning: Algorithms and Applications
Many real-world problems require trading off multiple competing objectives. However, these objectives are often in different units and/or scales, which can make it challenging for practitioners to express numerical preferences over objectives in their native units. In this talk I'll introduce algorithms for multi-objective reinforcement learning that enables setting desired preferences for objectives in a scale-invariant way. Furthermore, I will show applications of multi objective RL in different contexts including real and simulated robotics tasks.