BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Reinforcement Learning for 3D Molecular Design - Gregor Simm
DTSTART:20210405T153000Z
DTEND:20210405T163000Z
UID:TALK158689@talks.cam.ac.uk
CONTACT:Bingqing Cheng
DESCRIPTION:Automating molecular design using deep reinforcement learning 
 (RL) has the potential to greatly accelerate the search for novel material
 s. Despite recent progress on leveraging graph representations to design m
 olecules\, such methods are fundamentally limited by the lack of three-dim
 ensional (3D) information. To address this\, we present a novel RL formula
 tion for molecular design in Cartesian coordinates\, thereby extending the
  class of molecules that can be built. Our reward function is directly bas
 ed on fundamental physical properties such as the energy\, which we approx
 imate via fast quantum-chemical methods. To enable progress towards de-nov
 o molecular design\, we introduce MolGym\, an RL environment comprising se
 veral challenging molecular design tasks along with baselines. Further\, w
 e propose a novel actor-critic architecture that exploits the symmetries o
 f the design process through a rotationally covariant state-action represe
 ntation based on a spherical harmonics series expansion. In our experiment
 s\, we show that our agent can efficiently learn to solve molecular-design
  tasks from scratch that are unattainable with graph-based approaches.\n\n
 [1] G. N. C. Simm\, R. Pinsler\, J. M. Hernández-Lobato in Proceedings of
  the 37th International Conference on Machine Learning\, PMLR\, 2020\, pp.
  8959–8969\, http://proceedings.mlr.press/v119/simm20b.html.\n[2] G. N. 
 C. Simm\, R. Pinsler\, G. Csányi\, J. M. Hernández-Lobato in 9th Interna
 tional Conference on Learning Representations\, ICLR 2021\, https://openre
 view.net/forum?id=jEYKjPE1xYN.
LOCATION:virtual ZOOM meeting ID: 263 591 6003\, Passcode: 000042\, https:
 //us02web.zoom.us/j/2635916003?pwd=ZlBEQnRENGwxNmJGMENGMWxjak5nUT09
END:VEVENT
END:VCALENDAR