BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//University of California\, Berkeley//UCB Events Calendar//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
DTSTART:19701029T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:19700402T020000
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20170322T200257Z
DTSTART;TZID=America/Los_Angeles:20170323T170000
DTEND;TZID=America/Los_Angeles:20170323T180000
TRANSP:OPAQUE
SUMMARY:Student Applied Math Talk: Adaptive compression for eigenvalue problems
UID:108231-ucb-events-calendar@berkeley.edu
ORGANIZER;CN="UC Berkeley Calendar Network":
LOCATION:736 Evans Hall
DESCRIPTION:Michael Lindsey\, UC Berkeley\n\nLarge\, dense Hermitian eigenvalue problems play a central role in computational quantum chemistry. Often the matrices for these problems are of the form $A+B$\, where multiplication by $A$ is cheap\, multiplication by $B$ (and hence also by $A+B$) is costly\, and $\\Vert A \\Vert_2 \\gg \\Vert B \\Vert_2$. Although $B$ is less significant than $A$\, it is still too significant to be neglected\, and the evaluation of matrix-vector products $Bv$ typically constitutes the vast majority of the computational cost of standard iterative approaches. While $B$ itself cannot necessarily be approximated well by a low-rank matrix\, it can be compressed adaptively with respect to a low-dimensional subspace that is iteratively updated until convergence to the desired eigenspace is achieved. We will discuss the properties of the aforementioned â€˜adaptive compressionâ€™ operation\, as well as the convergence of the associated adaptive compression method for solving eigenvalue problems\, which has been adopted in community electronic structure software packages such as Quantum ESPRESSO. In particular\, we will explain how to prove local convergence with an asymptotic rate\, as well as global convergence that holds generically in a strong sense. The proof proceeds by studying the adaptive compression method as a dynamical system and ultimately takes some surprising turns through rather diverse fields of math.
URL:http://events.berkeley.edu/index.php/calendar/sn/pubaff.html?event_ID=108231&view=preview
SEQUENCE:0
CLASS:PUBLIC
CREATED:20170322T200257Z
LAST-MODIFIED:20170322T200257Z
X-MICROSOFT-CDO-BUSYSTATUS:BUSY
X-MICROSOFT-CDO-INSTTYPE:0
X-MICROSOFT-CDO-IMPORTANCE:1
X-MICROSOFT-CDO-OWNERAPPTID:-1
END:VEVENT
END:VCALENDAR