BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Rapid development of robust single purpose machine learning potent
 ials for complex aqueous systems - Dr Christoph Schran\, University of Cam
 bridge
DTSTART:20210203T143000Z
DTEND:20210203T153000Z
UID:TALK154189@talks.cam.ac.uk
CONTACT:Lisa Masters
DESCRIPTION:While machine learning has opened various new avenues in compu
 tational chemistry and material science\,\nit usually still remains a diff
 icult task to obtain robust and accurate models for a given system of inte
 rest.\nHere\, we show that committee neural network potentials can solve t
 his problem by providing readily developed\nsingle purpose models for comp
 lex molecular systems.\nUsing active learning based on query by committee 
 techniques\, a new model can be obtained in a one-step process\nfrom a sin
 gle reference trajectory.\nWe apply this methodology based on committee mo
 dels to multiple aqueous systems with increasing complexity.\nThe six aque
 ous systems chosen here comprise different ions in solution\, water in nan
 otubes and on titanium dioxide\ninterfaces\, as well as water under MoS2 c
 onfinement.\nHighlighting the accuracy of our approach\, the resulting mod
 els are validated in detail\nwith a new scoring scheme that includes struc
 tural and dynamical properties and\nthe precision of the force prediction 
 of the models.\nBy making the underlying packages available\, we think tha
 t such single purpose committee models\nwill enable the uncomplicated but 
 accurate extension of time and length scales in molecular simulations.
LOCATION:Zoom - Meeting ID: 922 6672 0201 Passcode: 427839
END:VEVENT
END:VCALENDAR