DrDavid Pfau

My primary research interests are in artificial intelligence and its applications to scientific problems, especially in computational physics. In collaboration with the Foulkes group at Imperial, I have focused on the use of deep learning in ab-initio electronic structure calculations, developing one of the first highly accurate neural network ansatzes for fermionic systems. We have continued to develop this work at Imperial, showing that these methods can be used to study quantum phase transitions, superfluidity, fractional quantum Hall states and other exotic condensed matter applications. In my primary role at Google DeepMind I have also worked on applications of AI to the control and optimization of plasmas for magnetic confinement fusion, and basic research on methods for "disentangling" different factors of variation in data manifolds. I believe that there are many rich connections between artificial intelligence, machine learning and computational physics, which require bridging the language between fields.