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:20211206T173455Z
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:70126406-0aee-452d-812b-3f52c3a84a79
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211206T173455Z
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:509e1236-9c51-4db5-aba5-91a819db68d8
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211206T173455Z
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:1ce380bc-bf7c-4d31-967e-7aa0fac14e25
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211206T173455Z
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:6e34dadf-7443-4c73-938f-58b77a36f6bc
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211206T173455Z
SUMMARY:Santanu Dey\, Title: Theoretical and computational analysis of si
zes of branch-and-bound trees
DESCRIPTION:Abstract: The branch-and-bound algorithm was invented for sol
ving integer programs (IP) in 1960. Since then\, there is almost no theo
retical analysis of the branch-and-bound algorithm\, even though the alg
orithm is the workhorse of all modern IP solvers. We try and answer some
of the following basic questions in this talk: (i) While it is known th
at the size of simple branch-and-bound trees can be exponential in size
in the worst case\, can we prove smaller size of branch-and-bound tree u
nder a random model for the instances? (ii) Most lower bounds on size of
branch-and-bound tree are for simple disjunctions. Can we prove similar
lower bounds for general disjunctions? (iii) Can we analyze the perform
ance of well-known branching rules like the full strong branching?\n\nTh
is is joint work with Yatharth Dubey\, Marco Molinaro\, and Prachi Shah.
\n\n
LOCATION:Zoom room 63658381373\; Seminar room 3721\, for watching the zo
om presentation on zoom together
DTSTART:20211105T130000Z
DTEND:20211105T140000Z
UID:ee927636-c96a-4cbe-b0a9-e521854a079d
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211206T173455Z
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:50397459-7de4-4fef-8961-49e830ff430b
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211206T173455Z
SUMMARY:Gabriele Eichfelder\, Multiobjective Mixed Integer Convex Optimiz
ation
DESCRIPTION:Abstract\nMultiobjective mixed integer convex optimization re
fers to mathematical programming problems where more than one convex obj
ective function needs to be optimized simultaneously and some of the var
iables are constrained to take integer values. In this talk\, we give a
short introduction to the basic concepts of multiobjective optimization.
We give insights why the famous approach of scalarization might not be
an appropriate method to solve these problems. Instead\, we present two
methods to solve the problems directly. The first is a branch-and-bound
method based on the use of properly defined lower bounds. We do not simp
ly rely on convex relaxations\, but we built linear outer approximations
of the image set in an adaptive way. The second method is purely based
on the criterion space. It uses ingredients from the well-known outer ap
proximation algorithm from single-objective mixed-integer optimization a
nd combines them with strategies to generate enclosures of nondominated
sets by iteratively improving approximations. For both algorithms\, we a
re able to guarantee correctness in terms of detecting the nondominated
set of multiobjective mixed integer convex problems according to a presc
ribed precision.\n
LOCATION:Hybrid: Seminar room 3721\, and Zoom room 63658381373
DTSTART:20211203T100000Z
DTEND:20211203T110000Z
UID:86eeb070-2096-4e6c-9451-4ef473b25683
END:VEVENT
BEGIN:VEVENT
DTSTAMP:20211206T173455Z
SUMMARY:Elina Rönnberg: Integer programming column generation
DESCRIPTION:Integer programming column generation: Accelerating branch-an
d-price using a novel pricing scheme for finding high-quality solutions
in set covering\, packing\, and partitioning problems\n
LOCATION:Seminar room 3721
DTSTART:20211210T100000Z
DTEND:20211210T110000Z
UID:f206013b-830b-469e-9548-84b81ff1187d
END:VEVENT
END:VCALENDAR