BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Machine learning in Physics\, Chemistry and Materi
 als discussion group (MLDG)
SUMMARY:Crystal Structure Search with Random Relaxations U
 sing Graph Networks - Gowoon Cheon\, Stanford
DTSTART;TZID=Europe/London:20210208T163000
DTEND;TZID=Europe/London:20210208T170000
UID:TALK157081AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/157081
DESCRIPTION:Materials design enables technologies critical to 
 humanity. While many properties of a material are 
 determined by its atomic crystal structure\, predi
 ction of the atomic crystal structure for a given 
 material's chemical formula is a long-standing gra
 nd challenge that remains a barrier in materials d
 esign. We build a novel dataset of more than 100\,
 000 random structure relaxations of battery anode 
 materials using high-throughput density functional
  theory (DFT) calculations\, which includes calcul
 ations with varying quantum mechanical settings fo
 r out-of-domain generalization. We modify graph ne
 ural network force fields to also predict stress i
 nformation\, which allows them to effectively simu
 late relaxations. We show that models trained on d
 ata conventionally used to train interatomic poten
 tials fail to simulate relaxations from random str
 uctures\, and random structure relaxations data is
  crucial for crystal structure search. We find tha
 t models trained with data augmentation via random
  perturbations improves both the accuracy and out 
 of domain generalization\, and is able to find an 
 experimentally verified structure of a new stoichi
 ometry.\n
LOCATION:virtual ZOOM meeting ID: 263 591 6003\, Passcode: 
 000042\, https://us02web.zoom.us/j/2635916003?pwd=
 ZlBEQnRENGwxNmJGMENGMWxjak5nUT09
CONTACT:Bingqing Cheng 
END:VEVENT
END:VCALENDAR