drew's talk

Author: dohyun-cse

_news/seminar_5.md

@@ -0,0 +1,19 @@
+---
+layout: post
+date: 2026-02-23 11:00:00-0500
+inline: true
+related_posts: false
+---
+
+#### _An Inexact Trust-Region Method for Nonsmooth PDE-Constrained Optimization_
+
+- **Speaker**: [Drew P. Kouri](https://www.sandia.gov/ccr/staff/drew-philip-kouri/) (Sandia National Laboratories)
+- **Date**: February 23, 2026
+- **Time**: 11:00 -- 12:00 (EST)
+- **Abstract**:
+  Optimization problems constrained by partial differential equations (PDEs) are ubiquitous in science and engineering, arising as optimal control, design and inverse problems. These problems are notoriously challenging to solve numerical. For example, simply evaluating the objective function requires solving a large-scale system of equations resulting from the discretized PDEs. This exorbitant cost necessitates the use of rapidly converging optimization routines to reduce the number of evaluations of the objective function and its derivatives. Unfortunately, this expense is exacerbated when the objective function involves nonsmooth terms such sparsifying regularizers. Traditional nonsmooth optimization methods converge (sub)linearly, often requiring many iterations to achieve marginal accuracy. In this talk, we discuss a proximal trust-region Newton method for minimizing the sum of a smooth nonconvex function and a nonsmooth convex function in Hilbert space. Our method is unique in that it permits and systematically controls inexact objective function and derivative evaluations. We prove global convergence of our method and establish, under mild assumptions, that it converges superlinearly, even quadratically. We demonstrate the efficiency of our algorithm on various examples from PDE-constrained optimization.
+- **Location**:
+  Virtual (Zoom)
+
+- **Zoom**: [Zoom Link](https://brown.zoom.us/j/9011251684?omn=97801700506&jst=2https://brown.zoom.us/j/9011251684?omn=97801700506&jst=2)
+- [**Add to Calendar**](/assets/calendar/seminar_5.ics)

assets/calendar/seminar_5.ics

@@ -0,0 +1,79 @@
+BEGIN:VCALENDAR
+CALSCALE:GREGORIAN
+VERSION:2.0
+X-WR-CALNAME:CIGMO Seminar Series - Drew P. Kouri
+METHOD:PUBLISH
+PRODID:-//Apple Inc.//macOS 15.6.1//EN
+BEGIN:VTIMEZONE
+TZID:America/New_York
+BEGIN:DAYLIGHT
+TZOFFSETFROM:-0500
+DTSTART:20070311T020000
+RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
+TZNAME:EDT
+TZOFFSETTO:-0400
+END:DAYLIGHT
+BEGIN:STANDARD
+TZOFFSETFROM:-0400
+DTSTART:20071104T020000
+RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
+TZNAME:EST
+TZOFFSETTO:-0500
+END:STANDARD
+END:VTIMEZONE
+BEGIN:VEVENT
+CREATED:20260213T155345Z
+DTEND;TZID=America/New_York:20260223T120000
+STATUS:CONFIRMED
+TRANSP:OPAQUE
+UID:7qqrn4og3ad6vkbgsu2gb55cvs@google.com
+DTSTART;TZID=America/New_York:20260223T110000
+LAST-MODIFIED:20260213T155357Z
+DTSTAMP:20260213T155602Z
+SUMMARY:CIGMO Seminar Series - Drew P. Kouri
+SEQUENCE:0
+DESCRIPTION:Speaker: Drew P. Kouri\, Sandia National Laboratories\nTitle
+ : An Inexact Trust-Region Method for Nonsmooth PDE-Constrained Optimizat
+ ion\n\nAbstract: Optimization problems constrained by partial differenti
+ al equations (PDEs) are ubiquitous in science and engineering\, arising 
+ as optimal control\, design and inverse problems. These problems are not
+ oriously challenging to solve numerical. For example\, simply evaluating
+  the objective function requires solving a large-scale system of equatio
+ ns resulting from the discretized PDEs. This exorbitant cost necessitate
+ s the use of rapidly converging optimization routines to reduce the numb
+ er of evaluations of the objective function and its derivatives. Unfortu
+ nately\, this expense is exacerbated when the objective function involve
+ s nonsmooth terms such sparsifying regularizers. Traditional nonsmooth o
+ ptimization methods converge (sub)linearly\, often requiring many iterat
+ ions to achieve marginal accuracy. In this talk\, we discuss a proximal 
+ trust-region Newton method for minimizing the sum of a smooth nonconvex 
+ function and a nonsmooth convex function in Hilbert space. Our method is
+  unique in that it permits and systematically controls inexact objective
+  function and derivative evaluations. We prove global convergence of our
+  method and establish\, under mild assumptions\, that it converges super
+ linearly\, even quadratically. We demonstrate the efficiency of our algo
+ rithm on various examples from PDE-constrained optimization.\n\n-::~:~::
+ ~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:
+ :~:~::-\nJoin Zoom Meeting\nhttps://brown.zoom.us/j/9011251684?omn=97801
+ 700506&jst=2 (ID: 9011251684)\n\nJoin by phone\n(US) +1 877-853-5247\n\n
+ Join using SIP\n9011251684@zoomcrc.com\n\nJoining instructions: https://
+ www.google.com/url?q=https://applications.zoom.us/addon/invitation/detai
+ l?meetingUuid%3DCzvek5oiTOuYTP7eoI8cRw%253D%253D%26signature%3Dfa57fc8ea
+ 75de99e257a20c61d4fcb0b00c4d43b297f9538946e31d12b43e817%26v%3D1&sa=D&sou
+ rce=calendar&usg=AOvVaw26UVIV1a4wEDewuclfJuXW\n\nMeeting host: <a href="
+ mailto:dohyun_kim@brown.edu" target="_blank">dohyun_kim@brown.edu</a><br
+  /><br />Join Zoom Meeting: <br /><a href="https://www.google.com/url?q=
+ https://brown.zoom.us/j/9011251684?omn%3D97801700506%26jst%3D2&amp\;sa=D
+ &amp\;source=calendar&amp\;usg=AOvVaw3d7zxMvl64Ss7nTt51vnax" target="_bl
+ ank">https://brown.zoom.us/j/9011251684?omn=97801700506&amp\;jst=2</a>\n
+ \nPlease do not edit this section.\n-::~:~::~:~:~:~:~:~:~:~:~:~:~:~:~:~:
+ ~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~:~::~:~::-
+BEGIN:VALARM
+UID:CDA06C1C-B882-442D-8CA4-C646DE997B04
+X-WR-ALARMUID:CDA06C1C-B882-442D-8CA4-C646DE997B04
+TRIGGER:-PT10M
+DESCRIPTION:This is an event reminder
+ACTION:DISPLAY
+END:VALARM
+END:VEVENT
+END:VCALENDAR