Nando de Freitas

About

Nando de Freitas is a distinguished machine learning researcher whose work spans Bayesian optimization, probabilistic inference, reinforcement learning, and robotics — fields in which he has made lasting and influential contributions. His early research established powerful probabilistic frameworks for sequential decision-making, including seminal work on Rao-Blackwellised particle filtering for dynamic Bayesian networks (145 citations) and active policy learning for robot exploration under uncertainty (148 citations). A recurring theme throughout his career is enabling intelligent agents to act effectively in complex, uncertain environments — from visually guided mobile robots (222 citations) to sophisticated manipulation systems trained end-to-end from raw camera inputs. His work on combining reinforcement learning with imitation learning for diverse visuomotor skills (217 citations) demonstrated how modest amounts of demonstration data can dramatically accelerate robotic learning. His contributions to Bayesian optimization, particularly scaling the technique to high-dimensional problems via random embeddings (242 citations), broadened its applicability across engineering and AI. More recently, de Freitas contributed to "Gato" (2022), DeepMind's ambitious generalist agent capable of performing hundreds of tasks across modalities and embodiments — a landmark step toward truly general artificial intelligence. With hundreds of citations across a decade of research, his impact on modern AI and robotics is profound and enduring.

Research Focus

Key Achievements

Top Papers

1. 1
   Bayesian optimization in high dimensions via random embeddings

   242 citations · 2013
2. 2
   A Bayesian exploration-exploitation approach for optimal online sensing and planning with a visually guided mobile robot

   222 citations · 2009
3. 3
   Reinforcement and Imitation Learning for Diverse Visuomotor Skills

   217 citations · 2018
4. 4
   Active Policy Learning for Robot Planning and Exploration under Uncertainty

   148 citations · 2007
5. 5
   Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks

   145 citations · 2013
6. 6
   Reinforcement and Imitation Learning for Diverse Visuomotor Skills

   116 citations · 2018
7. 7
   A Generalist Agent

   66 citations · 2022
8. 8
   Robust Imitation of Diverse Behaviors

   64 citations · 2017
9. 9
   Scaling data-driven robotics with reward sketching and batch reinforcement learning

   59 citations · 2020
10. 10
    Scaling data-driven robotics with reward sketching and batch reinforcement learning

    45 citations · 2019

Key Collaborators

Contact & Links