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:Theory of Condensed Matter
SUMMARY:Solving the Many-Electron SchrÃ¶dinger Equation wit
h Deep Neural Networks - Matthew Foulkes (Imperia
l College)
DTSTART;TZID=Europe/London:20200910T140000
DTEND;TZID=Europe/London:20200910T150000
UID:TALK150934AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/150934
DESCRIPTION:Exact wavefunctions of interesting many-electron s
ystems are NP-hard to compute in general\, but app
roximations can be found using polynomially-scalin
g algorithms. The challenge is to find an approxim
ate wavefunction that is simple enough to evaluate
and yet has enough variational freedom to produce
accurate results. Neural networks have shown impr
essive power as accurate practical function approx
imators and promise as compact wavefunctions for s
pin systems\, but the Pauli principle complicates
network representations of many-fermion wavefuncti
ons. Here we introduce a fully antisymmetric deep
learning architecture\, Fermi Net\, able to approx
imate the wavefunctions of atoms and small molecul
es to remarkable accuracy. For example\, we predic
t the dissociation curves of the nitrogen molecule
and the hydrogen chain\, two challenging strongly
-correlated systems\, to significantly higher prec
ision than the coupled cluster method\, widely con
sidered the best scalable method for quantum chemi
stry. This work opens the possibility of accurate
direct optimisation of wavefunctions for previousl
y intractable molecules and solids.
LOCATION:Details of video conferencing will be distributed
nearer the time.
CONTACT:Jan Behrends
END:VEVENT
END:VCALENDAR