BEGIN:VCALENDAR
PRODID:-//Ben Fortuna//iCal4j 1.0//EN
VERSION:2.0
CALSCALE:GREGORIAN
X-WR-CALNAME:Seminar\, Optimization and systems theory
BEGIN:VEVENT
DTSTAMP:20211023T050617Z
SUMMARY:Justin Pearson: The Essence of Constraint Programming
DESCRIPTION:Constraint Programming (CP) is a relatively young paradigm\,
geared\ntowards the elegant modelling and efficient solving of combinato
rial\nproblems\, which are so ubiquitous and important in management\,\n
engineering\, and science. CP works in a way orthogonal and\ncomplement
ary to other optimisation technologies\, such as integer\nprogramming (I
P)\, Boolean satisfiability (SAT)\, and answer-set\nprogramming (ASP).
CP has become the technology of choice in some\nareas\, such as scheduli
ng and configuration.\n\nI will present the essential principles of CP a
nd combinatorial\noptimisation\, and present some our research group's\n
(http://www.it.uu.se/research/group/optimisation research activities\nwi
thin CP and optimisation.\n
LOCATION:3418\, https://www.kth.se/places/room/id/774d2e00-417d-4642-b644
-f0c6b761e491
DTSTART:20210917T090000Z
DTEND:20210917T100000Z
UID:fe0d74bd-ad24-40ad-bd9e-67110f36d387
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211023T050617Z
SUMMARY:Julian Hall\, "HiGHS: Theory\, software and Impact"
DESCRIPTION:Abstract: Since Dantzig formulated the simplex algorithm in
1947\, the widespread need to solve linear optimization problems drove
the development of algorithmic and computational techniques for decades\
, yielding several high performance commercial and open source software
systems. This talk will focus on the Edinburgh-based work on solving lar
ge scale sparse linear programming problems that underpins the high perf
ormance open source linear optimization software\, HiGHS\, the challenge
s of developing such software\, and the Impact that it has achieved.\n
LOCATION:Seminar room 3418\, via zoom. (We show the presentation using th
e projector)
DTSTART:20211001T090000Z
DTEND:20211001T100000Z
UID:450266cf-82df-4644-bcd1-55bc78fd58ec
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211023T050617Z
SUMMARY:Yura Malitsky: Adaptive Gradient Descent without Descent
DESCRIPTION:Abstract: In this talk I will present some recent results for
the most classical optimization method — gradient descent. We will show
that a simple zero cost rule is sufficient to completely automate gradi
ent descent. The method adapts to the local geometry\, with convergence
guarantees depending only on the smoothness in a neighborhood of a solut
ion. The presentation is based on a joint work with K. Mishchenko\, see
https://arxiv.org/abs/1910.09529.\n
LOCATION:Seminar room 3721
DTSTART:20211015T090000Z
DTEND:20211015T100000Z
UID:95036fed-dbab-4cb9-93de-c7b78572b24d
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211023T050617Z
SUMMARY:Gonzalo Muñoz: Cutting planes for Non-Convex Quadratic Optimizati
on
DESCRIPTION:Abstract: The generation of tight approximations for Quadrati
cally Constrained Quadratic Programs (QCQPs) through strong valid linear
inequalities is an active and challenging research topic in the optimiz
ation community. Recently\, the generation of such inequalities for thes
e problems has been tackled by a number of authors using the intersectio
n cut paradigm - a highly studied tool in integer programming whose flex
ibility has triggered these renewed efforts in non-linear settings. In t
his talk\, we show how to construct intersection cuts in a quadratic set
ting using our proposed "maximal quadratic-free" sets. We describe the c
onstruction of these sets\, show how to compute valid inequalities from
them\, and evaluate this approach with extensive computational experimen
ts. This talk describes joint work with Antonia Chmiela and Felipe Serra
no.
LOCATION:Zoom room 63658381373
DTSTART:20211022T120000Z
DTEND:20211022T130000Z
UID:7130af3b-45bc-4359-bbce-111def63d7de
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211023T050617Z
SUMMARY:Santany Dey: TBA
DTSTART:20211105T130000Z
DTEND:20211105T140000Z
UID:472b5b0d-1819-4d0c-9c63-ff20fcd68efc
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211023T050617Z
SUMMARY:Dimitri Papageorgiou: Pooling problems under perfect and imperfec
t competition
DESCRIPTION:Abstract: We investigate pooling problems in which multiple p
layers vie with one another to maximize individual profit in a non-coope
rative competitive market. This competitive setting is interesting and w
orthy of study because the majority of prevailing process systems engine
ering models largely overlook the non-cooperative strategies that exist
in real-world markets. In this talk\, we provide a gentle overview of po
oling problems in which each player controls a processing network involv
ing intermediate tanks (or pools) where raw materials are blended togeth
er before being further combined into final products. Each player then s
olves a pure or mixed-integer bilinear optimization problem whose profit
is influenced by other players. We present several bilevel formulations
and numerical results of a novel decomposition algorithm. We demonstrat
e that our provably optimal decomposition algorithm can handle some of t
he largest bilevel optimization problems with nonconvex lower-level prob
lems ever considered in the literature.\n\nLink to arxiv preprint: http:
//arxiv.org/abs/2110.03018\n
LOCATION:Zoom room 63658381373
DTSTART:20211111T140000Z
DTEND:20211111T150000Z
UID:37ec4deb-6dd3-480c-a7e0-bceb6076da99
END:VEVENT
END:VCALENDAR