BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20241120T082410Z
LOCATION:HG F 30 Audi Max
DTSTART;TZID=Europe/Stockholm:20240603T185100
DTEND;TZID=Europe/Stockholm:20240603T185200
UID:submissions.pasc-conference.org_PASC24_sess156_pos113@linklings.com
SUMMARY:P02 - Accurate Machine Learning Force Fields via Experimental and 
 Simulation Data Fusion
DESCRIPTION:Poster\n\nSebastien Röcken and Julija Zavadlav (Technical Univ
 ersity of Munich)\n\nIn molecular dynamics, Machine Learning potentials (M
 LPs) have seen tremendous success when trained bottom-up on ab initio forc
 es and energies. MLPs enable simulation times out of reach for ab initio c
 omputations at accuracies out of reach for classical force fields. However
 , due to the underlying approximations when solving the Schrödinger equati
 on, MLPs sometimes fail to quantitatively reproduce experimental data. On 
 the other hand, training MLPs top-down on experimental target properties y
 ields largely under-constrained force fields that fail to reproduce many o
 ff-target properties for which bottom-up models yield better results. To o
 vercome these limitations, we present a combined bottom-up and top-down le
 arning approach using titanium as a showcase. We show that a fused trainin
 g approach yields an MLP in close agreement with DFT and experimental targ
 ets. Moreover, the fused model generalizes to several off-target propertie
 s and often performs better than training only on DFT data. The presented 
 approach is general and applicable to generate highly accurate MLPs for ot
 her materials.\n\nSession Chair: Erik W. Draeger (Lawrence Livermore Natio
 nal Laboratory)
END:VEVENT
END:VCALENDAR