BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Tim Hoheisel (McGill University)
DTSTART:20221013T223000Z
DTEND:20221013T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/1
DESCRIPTION:Title: The Maximum Entropy on the Mean Method for Linear Inverse Problems (a
nd beyond)\nby Tim Hoheisel (McGill University) as part of PIMS-CORDS
SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\
nThe principle of ‘maximum entropy’ states that the probability distri
bution which best represents the current state of knowledge about a system
is the one with largest entropy with respect to a given prior (data) dist
ribution. It was first formulated in the context of statistical physics in
two seminal papers by E. T. Jaynes (Physical Review\, Series II. 1957)\,
and thus constitutes an information theoretic manifestation of Occam’s r
azor. We bring the idea of maximum entropy to bear in the context of linea
r inverse problems in that we solve for the probability measure which is c
lose to the (learned or chosen) prior and whose expectation has small resi
dual with respect to the observation. Duality leads to tractable\, finite-
dimensional (dual) problems. A core tool\, which we then show to be useful
beyond the linear inverse problem setting\, is the ‘MEMM functional’:
it is an infimal projection of the Kullback- Leibler divergence and a lin
ear equation\, which coincides with Cramér’s function (ubiquitous in th
e theory of large deviations) in most cases\, and is paired in duality wit
h the cumulant generating function of the prior measure. Numerical example
s underline the efficacy of the presented framework.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nadia Lahrichi (Polytechnique Montreal)
DTSTART:20221027T223000Z
DTEND:20221027T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/2
DESCRIPTION:Title: Improving Radiotherapy Treatment Logistics\nby Nadia Lahrichi (Po
lytechnique Montreal) as part of PIMS-CORDS SFU Operations Research Semina
r\n\nLecture held in ASB 10908.\n\nAbstract\nThe main cancer treatments ar
e surgery\, radiation therapy and chemotherapy. The complexity of the logi
stical process of scheduling treatment appointments stems from the fact th
at it involves extremely costly resources\, sometimes synchronously. Sever
al due dates (i.e.\, appointments already scheduled\, maximum wait times)
and unexpected events such as the arrival of patients requiring urgent pal
liative care add to the difficulty. This talk will investigate how can sim
ulation and optimization models help improve the efficiency of cancer trea
tment centers and share experiences on patient booking\, physician schedul
ing\, and capacity assessment. All projects are conducted in close partner
ship with a hospital and rely on real data.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/2/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Imre Bárány (Rényi Institute and University College London)
DTSTART:20220929T223000Z
DTEND:20220929T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/3
DESCRIPTION:Title: Cells in the Box and a Hyperplane\nby Imre Bárány (Rényi Insti
tute and University College London) as part of PIMS-CORDS SFU Operations R
esearch Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nIt is well know
n that a line can intersect at most $2n-1$ cells of the $n \\times n$ ches
sboard. What happens in higher dimensions: how many cells of the $d$-dimen
sional $[0\,n]^d$ box can a hyperplane intersect? We also prove the intege
r analogue of the following fact. If $K\, L$ are convex bodies in $R^d$ an
d $K \\subset L$\, then the surface area $K$ is smaller than that of $L$.
Joint work with Peter Frankl.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jing Lu (University of Winchester)
DTSTART:20221117T233000Z
DTEND:20221118T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/4
DESCRIPTION:Title: Case Studies in Data Science and Analytics from a UK Business School<
/a>\nby Jing Lu (University of Winchester) as part of PIMS-CORDS SFU Opera
tions Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nData sci
ence involves the collection\, management\, processing\, analysis\, visual
isation and interpretation of huge amounts of data. It is a multi-discipli
nary field that integrates systematic thinking\, methodology\, process and
technology to develop intelligence with respect to real-world problems. T
his presentation focuses on the business environment and identifies the co
mponents of data science forming a conceptual architecture before proposin
g a composite data-driven process model. Representative tools and techniqu
es are applied to relevant case studies demonstrating innovation in underg
raduate programme design\, customer analytics and the marketing of insuran
ce for example.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/4/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Diego Cifuentes (remote) (Georgia Tech)
DTSTART:20221110T223000Z
DTEND:20221110T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/5
DESCRIPTION:Title: Computing the Nearest Structured Rank Deficient Matrix\nby Diego
Cifuentes (remote) (Georgia Tech) as part of PIMS-CORDS SFU Operations Res
earch Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nGiven an affine s
pace of matrices L and a matrix Θ ∈ L\, consider the problem of computi
ng the closest rank deficient matrix to Θ on L with respect to the Froben
ius norm. This is a nonconvex problem with several applications in control
theory\, computer algebra\, and computer vision. We introduce a novel sem
idefinite programming (SDP) relaxation\, and prove that it always gives th
e global minimizer of the nonconvex problem in the low noise regime\, i.e.
\, when Θ is close to be rank deficient. Our SDP is the first convex rela
xation for this problem with provable guarantees. We evaluate the performa
nce of our SDP relaxation in examples from system identification\, approxi
mate GCD\, triangulation\, and camera resectioning. Our relaxation reliabl
y obtains the global minimizer under non-adversarial noise\, and its noise
tolerance is significantly better than state of the art methods.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ben Adcock (Simon Fraser University)
DTSTART:20230126T233000Z
DTEND:20230127T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/8
DESCRIPTION:Title: Restarts Subject to Approximate Sharpness: a Parameter-free and Optim
al Scheme for Accelerating First-order Methods\nby Ben Adcock (Simon F
raser University) as part of PIMS-CORDS SFU Operations Research Seminar\n\
nLecture held in ASB 10908.\n\nAbstract\nSharpness is a generic assumption
in continuous optimization that bounds the distance to the set of minimi
zers in terms of the suboptimality in the objective function. It leads to
the acceleration of first-order optimization methods via so-called restar
ts. However\, sharpness involves problem-specific constants that are typi
cally unknown\, and previous restart schemes often result in reduced conv
ergence rates. Such schemes are also challenging to apply in the presence
of noise or approximate model classes (e.g.\, in compressed sensing or ma
chine learning problems). In this talk\, we introduce the notion of appro
ximate sharpness\, a generalization of sharpness that incorporates an unk
nown constant perturbation to the objective function error. By employing a
new type of search over the unknown constants\, we then describe a resta
rt scheme that applies to general first-order methods. Our scheme maintai
ns the same convergence rate as when assuming knowledge of the constants.
Moreover\, for broad classes of problems\, it gives rates of convergence
which either match known optimal rates or improve on previously establish
ed rates. Finally\, we demonstrate the practical efficacy of this scheme
on applications including sparse recovery\, compressive imaging and featu
re selection in machine learning.\n\nThis is joint work with Matthew J. Co
lbrook (Cambridge) and Maksym Neyra-Nesterenko (SFU). The corresponding p
aper can be found here: https://arxiv.org/abs/2301.02268\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Hossein Piri (University of Calgary)
DTSTART:20230105T233000Z
DTEND:20230106T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/9
DESCRIPTION:Title: Individualized Dynamic Patient Monitoring Under Alarm Fatigue\nby
Hossein Piri (University of Calgary) as part of PIMS-CORDS SFU Operations
Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nHospitals are
rife with alarms\, many of which are false. This leads to $alarm$ $fatig
ue$\, in which clinicians become desensitized and may inadvertently ignor
e real threats. We develop a partially observable Markov decision process
model for recommending dynamic\, patient-specific alarms in which we incor
porate a $cry$-$wolf$ feedback loop of repeated false alarms. Our model t
akes into account patient heterogeneity in safety limits for vital signs a
nd learns a patient’s safety limits by performing Bayesian updates durin
g a patient’s hospital stay. We develop structural results of the optima
l policy and perform a numerical case study based on clinical data from an
intensive care unit. We find that compared with current approaches of set
ting patients’ alarms\, our dynamic patient-centered model significantly
reduces the risk of patient harm.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sandy Rutherford (Simon Fraser University)
DTSTART:20230302T233000Z
DTEND:20230303T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/10
DESCRIPTION:Title: Simulation Modelling of the BC Critical Care System for Pandemic Res
ponse\nby Sandy Rutherford (Simon Fraser University) as part of PIMS-C
ORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbst
ract\nThe pandemic placed considerable stress on the critical care system
in British Columbia. In this talk\, I will present simulation modelling an
alysis done to support the response to the pandemic and ongoing work to im
prove the ability of the critical care system to respond to future public
health crises. The first project that I will discuss is a queuing model to
inform ventilator capacity planning during the first wave of the COVID-19
pandemic. I will then describe ongoing development of a discrete event si
mulation model for the network of major intensive care units (ICUs) in BC.
Currently\, our model contains admissions and transfers for ICUs and high
acuity units at eight hospitals in BC. This model will help to improve pa
tient access to critical care\, and inform planning for seasonal influenza
and COVID-19.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/10/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Tanmaya Karmarkar (UBC-O hosted) (UBC Okanagan)
DTSTART:20230119T233000Z
DTEND:20230120T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/11
DESCRIPTION:Title: Tensor Optimization and Applications\nby Tanmaya Karmarkar (UBC-
O hosted) (UBC Okanagan) as part of PIMS-CORDS SFU Operations Research Sem
inar\n\nLecture held in ASB 10908.\n\nAbstract\nFirst example of applying
tensor optimization to combinatorial problems was shown in IPCO 1992: page
s 406-420. We improve and strengthen those results in several ways and obt
ain computational results on three problems – graph partitioning\, satis
fiability and analysis of counterexamples related to Hilbert’s 17th prob
lem.\n\nFor this we created a mixed symbolic-numeric model formulation pac
kage which facilitates definition of objective function\, equality and ine
quality constraints and definition of new dependent variables.\n\nFor disc
rete problems certain inequalities valid at candidate solutions are dynami
cally incorporated in the iterations of the continuous optimization algori
thm based on underlying non-Newtonian geometry of the interior-point space
.\n\nFor graph partitioning we obtain optimal solutions including proof of
optimality. For satisfiability problem we either find the satisfiable ass
ignment or construct and output proof of unsatisfiability.\n\nFor Hilbert
’s 17th problem we analyse concrete examples whose non-negativity has be
en stablished to be not provable using sums of the squares expressions val
id in RN. However\, for these counterexamples\, we construct non-negativit
y proofs by computationally constructing sums of squares expressions valid
on certain sub-varieties of RN The same modeling package mentioned above
is used to post process the solver output into symbolic proofs of optimali
ty or infeasibility.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/11/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dimitri Leemans (Université Libre de Bruxelles)
DTSTART:20230202T233000Z
DTEND:20230203T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/12
DESCRIPTION:Title: The Number of String C-groups of High Rank\nby Dimitri Leemans (
Université Libre de Bruxelles) as part of PIMS-CORDS SFU Operations Resea
rch Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nAbstract polytopes
are a combinatorial generalisation of classical objects that were already
studied by the greeks. They consist in posets satisfying some extra axioms
. Their rank is roughly speaking the number of layers the poset has. When
they have the highest level of symmetry (namely the automorphism group has
one orbit on the set of maximal chains)\, they are called regular. One ca
n then use string C-groups to study them.\n\nIndeed\, string C-groups are
in one-to-one correspondence with abstract regular polytopes. They are als
o smooth quotients of Coxeter groups.\n\nThey consist in a pair $(G\,S)$ w
here $G$ is a group and $S$ is a set of generating involutions satisfying
a string property and an intersection property. The cardinality of the set
$S$ is the rank of the string C-group. It corresponds to the rank of the
associated polytope.\n\n \n\nIn this talk\, we will give the latest devel
opments on the study of string C-groups of high rank. In particular\, if $
G$ is a transitive group of degree $n$ having a string C-group of rank $rg
eq (n+3)/2$\, work over the last twelve years permitted us to show that $G
$ is necessarily the symmetric group $S_n$.\n\nWe have just proven in the
last months that if $n$ is large enough\, up to isomorphism and duality\,
the number of string C-groups of rank $r$ for $S_n$ (with $r \\geq (n+3)/2
$) is the same as the number of string C-groups of rank $r+1$ for $S_{n+1}
$. \n\nThis result and the tools used in its proof\, in particular the ra
nk and degree extension\, imply that if one knows the string C-groups of r
ank $(n+3)/2$ for $S_n$ with $n$ odd\, one can construct from them all str
ing C-groups of rank $(n+3)/2+k$ for $S_{n+k}$ for any positive integer $k
$. \n\nThe classification of the string C-groups of rank $r \\geq (n+3)/2
$ for $S_n$ is thus reduced to classifying string C-groups of rank $r$ for
$S_{2r-3}$.\n\nA consequence of this result is the complete classificatio
n of all string C-groups of $S_n$ with rank $n-\\kappa$ for $\\kappa \\in
{1\,\\ldots\,6}$\, when $n \\geq 2 \\kappa+3$\, which extends previous kn
own results.\n\nThe number of string C-groups of rank $n-\\kappa$\, with $
n \\geq 2 \\kappa +3$\, of this classification gives the following sequenc
e of integers indexed by $\\kappa$ and starting at $\\kappa = 1$.\n$$\\Sig
ma{\\kappa}=(1\,1\,7\,9\,35\,48).$$\nThis sequence of integers is new acco
rding to the On-Line Encyclopedia of Integer Sequences.\n\nJoint work with
Peter J. Cameron (University of St Andrews) and Maria Elisa Fernandes (Un
iversity of Aveiro)\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/12/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Gohram Baloch (Simon Fraser University)
DTSTART:20230316T223000Z
DTEND:20230316T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/13
DESCRIPTION:Title: Data-driven Approach to Optimal Ordering batching Problem in Warehou
se Management\nby Gohram Baloch (Simon Fraser University) as part of P
IMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\
nAbstract\nIn this work\, we focus on the picking process in warehouse man
agement and study it from a data perspective. Using historical data from a
n industrial partner\, we introduce\, model\, and study the robust order b
atching problem (ROBP) that groups orders into batches to minimize total o
rder processing time accounting for uncertainty caused by system congestio
n and human behavior. We provide a generalizable\, data-driven approach th
at overcomes warehouse-specific assumptions characterizing most of the wor
k in the literature. We analyze historical data to understand the processe
s in the warehouse\, to predict processing times\, and to improve order pr
ocessing. We introduce the ROBP and develop an efficient learning-based br
anch-and-price algorithm based on simultaneous column and row generation\,
embedded with alternative prediction models such as linear regression and
random forest that predict processing time of a batch. We conduct extensi
ve computational experiments to test the performance of the proposed appro
ach and to derive managerial insights based on real data.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/13/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Angela Morrison (UBCO-hosted) (University of Colorado - Denver)
DTSTART:20230216T233000Z
DTEND:20230217T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/14
DESCRIPTION:Title: On Combinatorial Algorithms and Circuit Augmentation for Pseudoflows
\nby Angela Morrison (UBCO-hosted) (University of Colorado - Denver) a
s part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in AS
B 10908.\n\nAbstract\nThere is a wealth of combinatorial algorithms for cl
assical min-cost flow\nproblems and their simpler variants like max flow o
r shortest-path problems. It is wellknown\nthat several of these algorithm
s are intimately related to the Simplex method\nand the more general circu
it augmentation schemes. Prime examples are the network\nSimplex method\,
a refinement of the primal Simplex method\, and (min-mean) cycle\ncancelin
g\, which corresponds to a (steepest-descent) circuit augmentation scheme
over\nthe underlying polyhedron.\n\nWe are interested in deepening and exp
anding the understanding of the close relationship\nbetween circuit augmen
tation and combinatorial network flows algorithms. To this end\,\nwe gener
alize from the consideration of primal or dual feasible flows to the so-ca
lled\npseudoflows\, which allow for a violation of flow balance. We introd
uce what are called\n‘pseudoflow polyhedra’\, in which slack variables
are used to quantify this violation\, and\ncharacterize their circuits. T
his enables us to study various network flows algorithms in\nview of the w
alks that they trace in these polyhedra\, and in view of the pivot rules u
sed\nto choose the steps.\n\nIn particular\, we show that the Successive S
hortest Path Algorithm and the Shortest/\nGeneric Augmenting Path Algorith
m form general\, non-edge circuit walks. We also\nprovide a proof outline
showing that the aforementioned algorithms correspond to a\nDantzig Rule a
nd Steepest-ascent circuit augmentation scheme respectively.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/14/
END:VEVENT
BEGIN:VEVENT
SUMMARY:[Reading break]
DTSTART:20230223T233000Z
DTEND:20230224T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/15
DESCRIPTION:Title: []\nby [Reading break] as part of PIMS-CORDS SFU Operations Rese
arch Seminar\n\nLecture held in ASB 10908.\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/15/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Isabelle Shankar (UBC-O hosted) (Portland State University)
DTSTART:20230525T223000Z
DTEND:20230525T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/16
DESCRIPTION:Title: Central Curve in Semidefinite Programming\nby Isabelle Shankar (
UBC-O hosted) (Portland State University) as part of PIMS-CORDS SFU Operat
ions Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThe Zaris
ki closure of the central path (which interior point algorithms track in c
onvex optimization\nproblems such as linear and semidefinite programs) is
an algebraic curve\, called the central curve. Its\ndegree has been studie
d in relation to the complexity of these interior point algorithms. We sho
w that\nthe degree of the central curve for generic semidefinite programs
is equal to the maximum likelihood\ndegree of linear concentration models.
This is joint work with Serkan Hosten and Angélica Torres.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/16/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mona Imanpoor Yourdshahy (Simon Fraser University)
DTSTART:20230323T223000Z
DTEND:20230323T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/18
DESCRIPTION:Title: Effects of Usage-Based Auto Insurance: A Dynamic Mechanism-Design Ap
proach\nby Mona Imanpoor Yourdshahy (Simon Fraser University) as part
of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908
.\n\nAbstract\nUsage-Based Insurance (UBI) is one of the most recent innov
ations by auto insurance companies that links the premium rates of custome
rs to their actual driving performance. In this program\, drivers’ behav
iours are monitored directly while they drive. Then\, the insurance compan
y uses this data to offer discounts on the insurance premium to their cust
omers. This paper provides a theoretical model to capture the effects of t
his monitoring technology on the auto insurance market. We formulate the u
nderlying insurance problem as a dynamic principal-agent model with hidden
information and hidden action. An agent (customer) privately knows his ty
pe that summarizes his ability as a driver and can exert an unobservable e
ffort in each period\, which affects his subsequent type. The principal (i
nsurer) offers a long-term contract to the agent despite the fact that she
observes neither the type of the agent nor the actions he takes. We chara
cterize the full history-dependent optimal contract for this dynamic adver
se selection and moral hazard problem. To compute the optimal contract\, w
e develop a general recursive formulation. The underlying system is a Mark
ov decision process\, where the evolution of the state of the system (type
of the customer) is endogenous\, as it depends on his hidden action in th
e previous period. We develop a dynamic programming algorithm to examine t
he model analytically and explore structural results about the optimal con
tract. The model results lead to important and interesting managerial insi
ghts for firms who may consider offering UBI programs. The study sheds lig
ht on how to design the contract to manage a UBI program\, the extent to w
hich a UBI policy can outperform a traditional policy\, and how the potent
ial gains depend on the demographics of the target market.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/18/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jean-François Cordeau (HEC Montréal)
DTSTART:20230310T233000Z
DTEND:20230311T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/19
DESCRIPTION:Title: The Park-and-loop Technician Routing Problem\nby Jean-François
Cordeau (HEC Montréal) as part of PIMS-CORDS SFU Operations Research Semi
nar\n\nLecture held in **SUR 2746**.\n\nAbstract\nMotivated by an applicat
ion in the routing of technicians at a\nFrench public utility\, we introdu
ce a highly efficient heuristic together\nwith a branch-price-and-cut algo
rithm for the doubly open park-and-loop\nrouting problem. This problem is
an extension of the classical vehicle\nrouting problem in which routes may
involve a main tour performed by driving\na vehicle as well as a set of s
ubtours that are carried out on foot after\nparking the vehicle. In additi
on\, routes do not start and end at a central\ndepot\, but rather at custo
mer locations. We first describe a matheuristic\nbased on a split procedur
e that generates high quality solutions fast. We\npresent computational ex
periments on a set of real instances with up to\n3\,800 customers. We also
apply the matheuristic to a related problem called\nthe vehicle routing p
roblem with transportable resources\, where the method\nfound new best sol
utions on 32 out of 40 benchmark instances from the\nliterature. We then p
resent an exact algorithm\, based on a set-covering\nformulation of the pr
oblem with columns representing complete routes\, which\nis capable of sol
ving to optimality instances with up to 50 customers.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/19/
END:VEVENT
BEGIN:VEVENT
SUMMARY:[Special talk on Friday in Surrey]
DTSTART:20230309T233000Z
DTEND:20230310T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/20
DESCRIPTION:Title: []\nby [Special talk on Friday in Surrey] as part of PIMS-CORDS
SFU Operations Research Seminar\n\nLecture held in ASB 10908.\nAbstract: T
BA\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/20/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Marco Caoduro (UBC Vancouver)
DTSTART:20230406T223000Z
DTEND:20230406T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/21
DESCRIPTION:Title: On the Packing and Hitting Numbers of Axis-parallel Segments\nby
Marco Caoduro (UBC Vancouver) as part of PIMS-CORDS SFU Operations Resear
ch Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nGiven a family R of
rectangles in the plane\, the packing number of R\, denoted by $\\nu$(R)\,
is the maximum size of a set of pairwise disjoint rectangles in R\, and t
he hitting number\, denoted by $\\tau$(R)\, is the minimum size of a set o
f points having a non-empty intersection with each rectangle in R. Clearly
\, $\\tau \\ge \\nu$.\n\nWegner (1965)\, and independently\, Gyárfás and
Lehel (1985)\, asked whether the hitting number $\\tau$ could be bounded
by a linear function of the packing number $\\nu$. In addition\, Wegner pr
oposed a multiplicative constant of 2. This problem is still wide open and
even if linear bounds are known for several particular cases\, almost non
e of them are paired with lower bound examples showing their optimality.\n
\nFor a family of axis-parallel line segments\, it is easy to show that $\
\tau \\le 2\\nu$\, as suggested by Wegner. During the talk\, we will consi
der families of axis-parallel segments with the additional property that n
o three of them meet at a point (that is\, the intersection graph is trian
gle-free). We show that\, in this restricted setting\, the packing number
of a family is at least $n/4+C_1\\sqrt{n}$ where $n$ is the size of the co
nsidered family and $C_1$ is a fixed positive constant. In addition\, we c
onstruct examples with packing number at most $n/4 + C_2\\sqrt{n}$ for a d
ifferent constant $C_2 > C_1$ showing that the previous bound is essential
ly optimal.\nAs a consequence of these results\, we settle the Wegner-Gyá
rfás-Lehel’s problem for axis-parallel segments showing that the multip
licative constant of 2 is optimal and deduce that $\\tau \\le 2\\nu − C_
3 \\sqrt{\\nu}$ for triangle-free axis-parallel segments. This bound canno
t be achieved for triangle-free axis-parallel rectangles\, marking a subst
antial difference in the behavior of segments and rectangles.\n\nAt the en
d of the talk\, we will present several open problems\, in particular\, li
nking our work with the recent developments on the computation of the pack
ing number for axis-parallel rectangles of Mitchell (2021) and Gálvez\, K
han\, Mari\, Mömke\, Pittu\, and Wiese (2022). This is joint work with Ja
na Cslovjecsek\, Michał Pilipczuk\, and Karol Węgrzycki.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/21/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Zhaosong Lu (Unniversity of Minnesota)
DTSTART:20230921T210000Z
DTEND:20230921T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/22
DESCRIPTION:Title: First-order Methods for Bilevel Optimization\nby Zhaosong Lu (Un
niversity of Minnesota) as part of PIMS-CORDS SFU Operations Research Semi
nar\n\nLecture held in ASB 10908.\n\nAbstract\nBilevel optimization has be
en widely used in a variety of areas such as adversarial training\, hyperp
arameter tuning\, image reconstruction meta-learning\, neural architecture
search\, and reinforcement learning. In this talk\, I will present first-
order methods for solving a class of bilevel optimization through the use
of single or sequential minimax optimization. The first-order operation co
mplexity of the proposed methods will be discussed.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/22/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sharan Vaswani (Simon Fraser University)
DTSTART:20231019T210000Z
DTEND:20231019T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/23
DESCRIPTION:Title: Exploiting Problem Structure for Efficient Optimization in Machine L
earning\nby Sharan Vaswani (Simon Fraser University) as part of PIMS-C
ORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbst
ract\nStochastic gradient descent (SGD) is the standard optimization metho
d for training machine learning (ML) models. SGD requires a step-size that
depends on unknown problem-dependent quantities\, and the choice of this
step-size heavily influences the algorithm's practical performance. By exp
loiting the interpolation property satisfied by over-parameterized ML mode
ls\, we design a stochastic line-search procedure that can automatically s
et the SGD step-size. The resulting algorithm exhibits improved theoretica
l and empirical convergence\, without requiring the knowledge of any probl
em-dependent constants. Next\, we consider efficient optimization for imit
ation learning (IL) and reinforcement learning. These settings involve opt
imizing functions for which it is expensive to compute the gradient. We pr
opose an optimization framework that uses the expensive gradient computati
on to construct surrogate functions that can then be minimized efficiently
. This allows for multiple model updates\, thus amortizing the cost of the
gradient computation. The resulting majorization-minimization algorithm i
s equipped with strong theoretical guarantees and exhibits fast convergenc
e on standard IL problems.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/23/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Julia Yan (UBC Sauder School of Business)
DTSTART:20231102T210000Z
DTEND:20231102T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/24
DESCRIPTION:Title: Pricing Shared Rides\nby Julia Yan (UBC Sauder School of Busines
s) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held i
n ASB 10908.\n\nAbstract\nShared rides\, which pool individual riders into
a single vehicle\, are essential for mitigating congestion and promoting
more sustainable urban transportation. However\, major ridesharing platfor
ms have long struggled to maintain a healthy and profitable shared rides p
roduct. To understand why shared rides have struggled\, we analyze procedu
res commonly used in practice to set static prices for shared rides\, and
discuss their pitfalls. We then propose a pricing policy that is adaptive
to matching outcomes\, dubbed match-based pricing\, which varies prices de
pending on whether a rider is dispatched alone or to what extent she is ma
tched with another rider. Analysis on a single origin-destination setting
reveals that match-based pricing is both profit-maximizing and altruistic\
, simultaneously improving cost efficiency (i.e.\, the fraction of cost sa
ved by shared rides relative to individual rides) and reducing rider payme
nts relative to the optimal static pricing policy. These theoretical resul
ts are validated on a large-scale simulation with hundreds of origin-desti
nations from Chicago ridesharing data. The improvements in efficiency and
reductions in payments are especially noticeable when costs are high and d
emand density is low\, enabling healthy operations where they have histori
cally been most challenging.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/24/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Heinz Bauschke (UBC-O hosted) (UBC Okanagan)
DTSTART:20231005T210000Z
DTEND:20231005T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/25
DESCRIPTION:Title: On the Composition of Two Linear Projections\nby Heinz Bauschke
(UBC-O hosted) (UBC Okanagan) as part of PIMS-CORDS SFU Operations Researc
h Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nProjection operators
are fundamental algorithmic operators in Analysis and Optimization. It is
well known that these operators are firmly nonexpansive\; however\, their
composition is generally only averaged and no longer firmly nonexpansive
. We will introduce the modulus of averagedness and provide an exact resul
t for the composition of two linear projection operators. As a consequence
\, we deduce that the Ogura-Yamada bound for the modulus of the compositio
n is sharp. Based on joint work with Theo Bendit and Walaa Moursi.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/25/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Feyza Sahinyazan (SFU Beedie School of Business)
DTSTART:20231116T220000Z
DTEND:20231116T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/26
DESCRIPTION:Title: Path to Energy Sovereignty: Clean and Affordable Solutions for Remot
e Communities\nby Feyza Sahinyazan (SFU Beedie School of Business) as
part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB
10908.\n\nAbstract\nRemote communities around the globe rely on off-grid i
nstallations of stand-alone diesel generators to cover their energy needs\
, which can be costly\, harmful to the environment and subject to disrupti
ons. Policymakers seek sustainable solutions for these communities to meet
the Sustainable Development Goals regarding clean energy and reduced ineq
ualities. Even with the best intentions\, ignoring community perspectives
can hamper the clean energy transition and energy accessibility goals of r
emote communities. Our objective in this research is to identify the optim
al generation capacity investment decisions from a remote community’s pe
rspective and investigate how common policy mechanisms interact with these
decisions.\n\n\nI am also planning to dedicate some portion of my talk to
give a brief overview of my other ongoing projects to see if there is any
interest in collaboration.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/26/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Frederik Kunstner (UBC)
DTSTART:20231130T220000Z
DTEND:20231130T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/27
DESCRIPTION:Title: Searching for Optimal Per-Coordinate Step-sizes with Multidimensiona
l Backtracking\nby Frederik Kunstner (UBC) as part of PIMS-CORDS SFU O
perations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThe
backtracking line-search is an effective technique to automatically tune t
he step-size in smooth optimization. It guarantees similar performance to
using the theoretically optimal step-size. Many approaches have been devel
oped to instead tune per-coordinate step-sizes\, also known as diagonal pr
econditioners\, but none of the existing methods are provably competitive
with the optimal per-coordinate stepsizes. We propose multidimensional bac
ktracking\, an extension of the backtracking line-search to find good diag
onal preconditioners for smooth convex problems. Our key insight is that t
he gradient with respect to the step-sizes\, also known as hypergradients\
, yields separating hyperplanes that let us search for good preconditioner
s using cutting-plane methods. As black-box cutting-plane approaches like
the ellipsoid method are computationally prohibitive\, we develop an effic
ient algorithm tailored to our setting. Multidimensional backtracking is p
rovably competitive with the best diagonal preconditioner and requires no
manual tuning.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/27/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Eitan Levin (UBC-O hosted) (California Institute of Technology)
DTSTART:20240125T220000Z
DTEND:20240125T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/28
DESCRIPTION:Title: Any-dimensional Convex Sets\nby Eitan Levin (UBC-O hosted) (Cali
fornia Institute of Technology) as part of PIMS-CORDS SFU Operations Resea
rch Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nClassical algorithm
s are defined on inputs of different sizes. In contrast\, data-driven algo
rithms\, that is\, algorithms learned from some data\, may only be defined
on inputs of the same size as the data. What\ndoes it mean for an algorit
hm to be defined on infinitely-many input sizes? How do we describe such\n
algorithms\, and how do we parametrize and search over them?\n\nIn this ta
lk\, we tackle these questions for convex optimization-based algorithms. D
escribing such\nalgorithms reduces to describing convex sets. These\, in t
urn\, are often "freely" described\, meaning that their description makes
instantiation in every dimension obvious. Examples include unit balls of\n
standard norms defined on vectors of any size\, graph parameters defined f
or graphs of any size\, and\n(quantum) information theoretic quantities de
fined for distributions on any number of (qu)bits.\n\nWe show that such fr
ee descriptions of convex sets arise from two ingredients. First\, group i
nvariance\nand the recently-identified phenomenon of representation stabil
ity. Second\, embeddings and projections\nrelating different-sized problem
instances. We combine these ingredients to obtain parametrized\nfamilies
of infinitely instantiable convex sets. To extend a set learned from data
in a fixed dimension to higher ones\, we identify consistency conditions r
elating sets in different dimensions that are satisfied in a variety of ap
plications\, and obtain parametrizations respecting these conditions. Our
parametrizations can be obtained computationally.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/28/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Warren Hare (UBC-O hosted) (UBC Okanagan)
DTSTART:20240111T220000Z
DTEND:20240111T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/29
DESCRIPTION:Title: Expected Decrease for Derivative-free Algorithms Using Random Subspa
ces\nby Warren Hare (UBC-O hosted) (UBC Okanagan) as part of PIMS-CORD
S SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstrac
t\nDerivative-free algorithms seek the minimum of a given function based o
nly on function values queried at appropriate points. Their performance is
known to worsen as the problem dimension increases.\n \nRecent advances
in developing randomized derivative-free techniques have tackled this iss
ue by working in low-dimensional subspaces that are drawn at random in an
iterative fashion. In this talk\, we present analysis for derivative-free
algorithms that employing random subspaces to obtain understanding of the
expected decrease achieved per function evaluation.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/29/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Charles Audet (UBC-O hosted) (Polytechnique Montréal)
DTSTART:20240314T210000Z
DTEND:20240314T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/30
DESCRIPTION:Title: Evolution of the Mads Algorithm by Developing Specific Features\
nby Charles Audet (UBC-O hosted) (Polytechnique Montréal) as part of PIMS
-CORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAb
stract\nThe topic of this talk is on Blackbox optimization (BBO)\, the stu
dy of\napplications\, attributes\, and solutions of optimization problems
in\nwhich the values of one or more of the functions defining the problem\
nare provided through blackboxes. The Mesh Adaptive Direct Search (Mads)\n
is a derivative-free algorithm pioneered in 2006 for constrained BBO\nprob
lems. This talk discusses recent advances to the Mads algorithm\,\nincludi
ng -i- the treatment of granular variables\; -ii- dynamic scaling\nof vari
ables\; -iii- escaping unknown discontinuities\; and -iv- revised\nconverg
ence results for discontinuous functions.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/30/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yves Lucet (UBC-O hosted) (UBC-Okanagan)
DTSTART:20240229T220000Z
DTEND:20240229T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/31
DESCRIPTION:Title: Recent Results in Computational Convex Analysis\nby Yves Lucet (
UBC-O hosted) (UBC-Okanagan) as part of PIMS-CORDS SFU Operations Research
Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nComputational convex a
nalysis aims at computing mathematical objects commonly used in convex ana
lysis with an emphasis on lower dimensions to facilitate visualization. Th
e latest contributions have focused on computing the closest convex functi
on and the ongoing quest for explicit formulas for the conjugate of noncon
vex functions. To ease the development of a toolbox\, greater emphasis is
put on simplifying intermediate computations especially for piecewise func
tions because the overall computation time depends on the number of pieces
.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/31/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yiwen Chen (UBC-O hosted) (UBC Okanagan)
DTSTART:20240404T210000Z
DTEND:20240404T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/32
DESCRIPTION:Title: Q-fully quadratic modeling and its application in a random subspace
derivative-free method\nby Yiwen Chen (UBC-O hosted) (UBC Okanagan) as
part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB
10908.\n\nAbstract\nDerivative-free optimization (DFO) methods are a clas
s of optimization methods that do not use the derivatives of the objective
or constraint functions. Model-based DFO methods are an important class
of DFO methods that are known to struggle with solving high-dimensional op
timization problems. Recent research has shown that incorporating random
subspaces into model-based DFO methods has the potential to improve their
performance on high-dimensional problems. However\, most of the current th
eoretical and practical results are based on linear approximation models d
ue to the complexity of quadratic approximation models. In this talk\, we
propose a random subspace derivative-free trust-region algorithm based on
quadratic approximations. Unlike most of its precursors\, this algorithm d
oes not require any special form of objective function. We study the geome
try of sample sets\, the error bounds for approximations\, and the quality
of subspaces. In particular\, we provide a technique to construct Q-fully
quadratic models\, which is easy to analyze and implement. We present an
almost-sure global convergence result of our algorithm and give an upper b
ound on the expected number of iterations to find a sufficiently small gra
dient. We also develop numerical experiments to compare the performance of
our algorithm using both linear and quadratic approximation models. The n
umerical results demonstrate the strengths and weaknesses of using quadrat
ic approximations.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/32/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Fatemeh Beik (Vali-e-Asr University of Rafsanjan)
DTSTART:20240509T210000Z
DTEND:20240509T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/33
DESCRIPTION:Title: A survey on preconditioning techniques for a class of block three-by
-three linear systems\nby Fatemeh Beik (Vali-e-Asr University of Rafsa
njan) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture hel
d in ASB 10908.\n\nAbstract\nIn this talk\, we study the performance of so
me preconditioners for accelerating the convergence of Krylov subspace met
hods for solving linear systems of equations with a block three-by-three s
tructure. A brief discussion is included regarding how spectral and field-
of-value analyses can be exploited to study the performance of a precondit
ioner in conjunction with the Generalized Minimum Residual Method (GMRES).
Numerical experiments show the effectiveness of inexact versions of preco
nditioners used with flexible GMRES for solving linear systems of equation
s arising from mixed finite element discretizations of the coupled Stokes-
Darcy flow problem.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/33/
END:VEVENT
BEGIN:VEVENT
SUMMARY:João Gouveia (UBC-O hosted) (University of Coimbra)
DTSTART:20240425T210000Z
DTEND:20240425T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/34
DESCRIPTION:Title: Self-dual polyhedral cones and their slack matrices\nby João Go
uveia (UBC-O hosted) (University of Coimbra) as part of PIMS-CORDS SFU Ope
rations Research Seminar\n\nLecture held in ASB 10908.\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/34/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Benny Wai (Pattison Food Group Ltd.)
DTSTART:20240418T210000Z
DTEND:20240418T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/35
DESCRIPTION:Title: Data Science Initiatives in the Grocery Industry\nby Benny Wai (
Pattison Food Group Ltd.) as part of PIMS-CORDS SFU Operations Research Se
minar\n\nLecture held in ASB 10908.\n\nAbstract\nPattison Food Group Ltd.\
, encompassing Save-On-Food\, Nester’s Market\, and over a dozen other g
rocery brands\, stands as a leading provider of food and health products i
n Western Canada. In this presentation\, Benny Wai\, Data & Analytics Mana
ger at PFG and SFU alumnus\, will delve into the machine learning and opti
mization solutions currently implemented at PFG. We will also navigate the
complexities and growing pains associated with establishing a new D&A dep
artment within an organization with over a century of history\, highlighti
ng the transformative journey towards data-driven decision-making.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/35/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yohan Song (University of Waterloo)
DTSTART:20240620T213000Z
DTEND:20240620T223000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/36
DESCRIPTION:Title: A Combinatorial Puzzle of Skew Shapes\nby Yohan Song (University
of Waterloo) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLec
ture held in ASB 10908.\n\nAbstract\nA skew shape is a difference of two Y
oung diagrams where one diagram contains the other. In 2017\, Jenna Rajchg
ot\, Matthew Satriano\, and Wanchun Shen showed that skew shapes can be us
ed to study the Gerstenhaber problem\, a matrix algebra problem in commuta
tive algebra. In this talk\, I present a recent progress with this method\
, albeit the approach is purely combinatorial.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/36/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Curtis Bright (University of Windsor)
DTSTART:20240822T210000Z
DTEND:20240822T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/37
DESCRIPTION:Title: SAT and Lattice Reduction for Integer Factorization\nby Curtis B
right (University of Windsor) as part of PIMS-CORDS SFU Operations Researc
h Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThe difficulty of fac
toring large integers into primes is the basis for cryptosystems such as R
SA. Due to the widespread popularity of RSA\, there have been many propose
d attacks on the factorization problem such as side-channel attacks where
some bits of the prime factors are available. When enough bits of the prim
e factors are known\, two methods that are effective at solving the factor
ization problem are satisfiability (SAT) solvers and Coppersmith's method.
The SAT approach reduces the factorization problem to a Boolean satisfiab
ility problem\, while Coppersmith's approach uses lattice basis reduction.
Both methods have their advantages\, but they also have their limitations
: Coppersmith's method does not apply when the known bit positions are ran
domized\, while SAT-based methods can take advantage of known bits in arbi
trary locations\, but have no knowledge of the algebraic structure exploit
ed by Coppersmith's method. In this paper we describe a new hybrid SAT and
computer algebra approach to efficiently solve random leaked-bit factoriz
ation problems. Specifically\, Coppersmith's method is invoked by a SAT so
lver to determine whether a partial bit assignment can be extended to a co
mplete assignment. Our hybrid implementation solves random leaked-bit fact
orization problems significantly faster than either a pure SAT or pure com
puter algebra approach.\n\nThis is joint work with Yameen Ajani.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/37/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ahmadreza Marandi (UBC-O hosted) (Eindhoven University)
DTSTART:20240828T210000Z
DTEND:20240828T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/38
DESCRIPTION:Title: A Clustering-based uncertainty set for Robust Optimization\nby A
hmadreza Marandi (UBC-O hosted) (Eindhoven University) as part of PIMS-COR
DS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstra
ct\nRobust Optimization (RO) is an approach to tackle uncertainties in the
parameters of an optimization\nproblem. Constructing an uncertainty set i
s crucial for RO\, as it determines the quality and the\nconservativeness
of the solutions. In this talk\, we introduce an approach for constructing
a data-driven\nuncertainty set through volume-based clustering\, which we
call Minimum-Volume Norm-Based\nClustering (MVNBC)\, that leads to less c
onservative solutions. MVNBC extends the concept of\nMinimum-Volume Ellips
oid Clustering by enabling customizable regions containing clusters. These
\nregions are defined based on a given set of vector norms\, hence providi
ng great flexibility in capturing\ndiverse data patterns. We formulate a m
ixed-integer conic optimization problem for MVNBC. To\naddress computation
al complexities\, we design an efficient iterative approximation algorithm
where we\nreassign points to clusters and improve the volume of the regio
ns. Our numerical experiments\ndemonstrate the effectiveness of our approa
ch in capturing data patterns and finding clusters with\nminimum total vol
ume. Moreover\, constructed uncertainty sets based on MVNBC result in robu
st\nsolutions with 10% improvement in the objective value compared to the
ones obtained by a recent datadriven\nuncertainty set. Therefore\, using o
ur uncertainty sets in RO problems can generate less\nconservative solutio
ns compared to traditional uncertainty sets as well as other existing data
-driven\napproaches.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/38/
END:VEVENT
BEGIN:VEVENT
SUMMARY:16 speakers (Held at UBC Vancouver)
DTSTART:20240921T153000Z
DTEND:20240921T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/39
DESCRIPTION:Title: 2024 West Coast Optimization Meeting\nby 16 speakers (Held at UB
C Vancouver) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLect
ure held in 110 Hugh Dempster Pavilion.\n\nAbstract\nDetails at: \n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/39/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Gabriel Jarry-Bolduc (UBC-O hosted) (Mount Royal University)
DTSTART:20241017T210000Z
DTEND:20241017T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/40
DESCRIPTION:Title: The cosine measure relative to a subspace\nby Gabriel Jarry-Bold
uc (UBC-O hosted) (Mount Royal University) as part of PIMS-CORDS SFU Opera
tions Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThe cosi
ne measure is a tool that tells you how well a set of directions is coveri
ng the space $\\mathbb{R}^n$. In this talk\, we extend the concept of cosi
ne measure by defining the cosine measure relative to a subspace.\nThis no
vel definition might be useful for subspace decomposition optimization met
hods. We propose a\ndeterministic algorithm to compute it and discuss the
situation in which the set of directions is infinite.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/40/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Shawn Wang (UBC-O hosted) (UBC-Okanagan)
DTSTART:20241003T210000Z
DTEND:20241003T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/41
DESCRIPTION:Title: On Bauschke-Bendit-Moursi modulus of averagedness\nby Shawn Wang
(UBC-O hosted) (UBC-Okanagan) as part of PIMS-CORDS SFU Operations Resear
ch Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nFirmly nonexpansive
operators are important in Convex Analysis and Optimization and Algorithms
. It is a special case of averaged operators. We classify averaged operato
rs\, firmly nonexpansive operators\, and proximal mappings by\nthe BBW mod
ulus of averagedness. One amazing result is that a proximal mapping of a c
onvex function has its modulus of averagedness less than $1/2$ if and only
if the function is Lipschitz smooth. Joint work with Shuang Song.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/41/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nitya Mani (UBC-O hosted) (MIT)
DTSTART:20241107T220000Z
DTEND:20241107T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/42
DESCRIPTION:Title: Tetrahedron intersecting families of hypergraphs\nby Nitya Mani
(UBC-O hosted) (MIT) as part of PIMS-CORDS SFU Operations Research Seminar
\n\nLecture held in ASB 10908.\n\nAbstract\nAn $H$-intersecting family of
3-uniform hypergraphs on $n$ labelled vertices is a family of hypergraphs
$\\mathcal{F}$ such that for any pair of hypergraphs $G_1\, G_2 \\in \\mat
hcal{F}$\, the intersection $G_1 \\cap G_2$ contains a copy of $H$ as a su
bgraph. One can construct a large such family $\\mathcal{F}$ by choosing a
ll of the hypergraphs that contain a fixed copy of $H$\, a family with siz
e $2^{{n \\choose 3} - e(H)}$. Understanding for which cases such a family
is asymptotically maximal is a very old and well-studied question\, and i
t has been conjectured that this lower bound is tight whenever $H$ is a co
mplete graph. The case of triangle intersecting families of graphs was stu
died by Shearer and was one of the first application's of Shearer’s entr
opy inequality to a combinatorial problem. This triangle-intersecting prob
lem was fully resolved by Ellis\, Filmus\, and Friedgut\, and more recentl
y the case of $K_4$-intersecting graphs was resolved by Berger and Zhao\,
both using linear programming bounds. Despite this progress\, understandin
g the maximal size of an $H$-intersecting family remains open for every ot
her complete (hyper)graph. In join work with Owen Zhang\, we resolve the c
ase of $K_5$-intersecting families and provide the first resolution of an
instance in the hypergraph setting\, showing that the conjecture holds for
tetrahedron-intersecting families of 3-uniform hypergraphs.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/42/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Betty Shea (UBC Vancouver)
DTSTART:20241114T220000Z
DTEND:20241114T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/43
DESCRIPTION:Title: Why Line-Search When You Can Plane-Search?\nby Betty Shea (UBC V
ancouver) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture
held in ASB 10908.\n\nAbstract\nThe practical performance of an optimizat
ion method depends on details such as using good step sizes. Strategies fo
r setting step sizes are generally limited to hyperparameter tuning (for a
fixed step size)\, step size schedules and line searches. For many common
machine learning problems\, line optimization and subspace optimization f
ind accurate step sizes for asymptotically the same cost as using a fixed
step size. In some cases\, line optimization may find step sizes that are
ruled out by the standard Armijo condition. For optimization methods that
use multiple search directions\, such as gradient descent with momentum\,
using subspace optimization instead of fixed step size schedules allow for
better adaptivity and potentially faster convergence. In the case of some
neural networks\, subspace optimization allows the use of different step
sizes for different layers that could decrease the amount of training time
needed\, as well as reducing the dependence on hyperparameter tuning.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/43/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sandy Rutherford (SFU)
DTSTART:20241128T220000Z
DTEND:20241128T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/44
DESCRIPTION:Title: Critical Care Planning for Pandemic Response\nby Sandy Rutherfor
d (SFU) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture h
eld in ASB 10908.\n\nAbstract\nThe COVID-19 Pandemic placed considerable s
train on intensive care units\, and the critical care system in British Co
lumbia and worldwide. I will review the critical care system and how it re
sponded to the COVID-19 pandemic. During the first wave of the COVID-19 Pa
ndemic\, we developed a simulation model to inform mechanical ventilator a
ccess in BC. One of the challenges that we faced is that simulation models
are difficult to study under epidemic growth in demand. I will describe a
pproximation methods from queueing theory that we used to address this cha
llenge. Specifically\, I will explore the accuracy of three queueing theor
y approximations under epidemic growth in demand\, namely: the pointwise s
tationary approximation\, the modified offered load approximation\, and th
e fixed-point approximation. We found that the fixed-point approximation i
s the most accurate and a hybrid optimization approach combining the fixed
-point approximation with simulation optimization was developed to determi
ne the number of mechanical ventilators required to ensure that at least 9
5% of patients could access a ventilator immediately during the first wave
of the COVOD-19 pandemic. Strengthening the BC critical care system to re
spond to seasonal respiratory illnesses and future pandemics is a priority
of the Ministry of Health. I will describe a large-scale simulation model
that we have developed to support this effort and discuss how operations
research can contribute to improving quality of care\, efficiency\, and re
siliency in the critical care system.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/44/
END:VEVENT
BEGIN:VEVENT
SUMMARY:[Halloween]
DTSTART:20241031T210000Z
DTEND:20241031T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/45
DESCRIPTION:by [Halloween] as part of PIMS-CORDS SFU Operations Research S
eminar\n\nLecture held in ASB 10908.\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/45/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jiajin Li (UBC Sauder)
DTSTART:20250123T220000Z
DTEND:20250123T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/46
DESCRIPTION:Title: Unveiling Spurious Stationarity and Hardness Results for Bregman Pro
ximal-Type Algorithms\nby Jiajin Li (UBC Sauder) as part of PIMS-CORDS
SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract
\nBregman proximal-type algorithms\, such as mirror descent\, are popular
in optimization and data science for effectively exploiting problem struct
ures and optimizing them under tailored geometries. However\, most of exis
ting convergence results rely on the gradient Lipschitz continuity of the
kernel\, which unfortunately excludes most commonly used cases\, such as t
he Shannon entropy. In this paper\, we reveal a fundamental limitation of
these methods: Spurious stationary points inevitably arise when the kern
el is not gradient Lipschitz. The existence of these spurious stationary p
oints leads to an algorithm-dependent hardness result: Bregman proximal-ty
pe algorithms cannot escape from a spurious stationary point within any fi
nite number of iterations when initialized from that point\, even in conve
x settings. This limitation is discovered through the lack of a well-defin
ed stationarity measure based on Bregman divergence for non-gradient Lipsc
hitz kernels. Although some extensions attempt to address this issue\, we
demonstrate that they still fail to reliably distinguish between stationar
y and non-stationary points for such kernels. Our findings underscore the
need for new theoretical tools and algorithms in Bregman geometry\, paving
the way for further research.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/46/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Serhii Myroshnychenko (University of the Fraser Valley)
DTSTART:20241024T210000Z
DTEND:20241024T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/47
DESCRIPTION:Title: Centroid of a convex body can be rarely the centroid of its section
s\nby Serhii Myroshnychenko (University of the Fraser Valley) as part
of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908
.\n\nAbstract\nWe construct a convex body $K$ in $R^n$\, $n \\ge 5$\, with
the property that there is exactly one hyperplane $H$ passing through $c(
K)$\, the centroid of $K$\, such that the centroid of $K \\cap H$ coincide
s with $c(K)$. This provides answers to questions of Grunbaum and Loewner
for $n \\ge 5$. The proof is based on the existence of non-intersection bo
dies in these dimensions. Joint work with K. Tatarko and V. Yaskin.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/47/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Kevin Hare (UBC-O hosted) (University of Waterloo)
DTSTART:20241205T220000Z
DTEND:20241205T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/48
DESCRIPTION:Title: Computational Progress on the Unfair 0-1 Polynomial Conjecture\n
by Kevin Hare (UBC-O hosted) (University of Waterloo) as part of PIMS-CORD
S SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstrac
t\nLet $c(x)$ be a monic integer polynomial with coefficients 0 or 1. Writ
e\n$c(x)=a(x)b(x)$ where $a(x)$ and $b(x)$ are monic polynomials with\nnon
-negative real (not necessarily integer) coefficients. The unfair\n0-1 pol
ynomial conjecture states that $a(x)$ and $b(x)$ are necessarily\ninteger
polynomials with coefficients 0 or 1.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/48/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Krishna Narayanan (SFU)
DTSTART:20241212T220000Z
DTEND:20241212T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/49
DESCRIPTION:Title: Coping with Intractability: Parameterized Algorithms meets Linear Pr
ogramming\nby Krishna Narayanan (SFU) as part of PIMS-CORDS SFU Operat
ions Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThis semi
nar will highlight the role of linear programming techniques in the design
of parametrized algorithms as a framework to cope with intractability\, w
hich I will attempt to motivate. After a brief introduction\, I will also
briefly talk about the associated notion of kernelization\, which transfor
ms input instances into more “manageable forms”. I will then demonstra
te how linear programming is applied in lieu of this framework to obtain a
reasonable algorithm from a parametrized complexity perspective for an ot
herwise intractable problem like vertex cover. Lastly\, recent advancement
s that use similar techniques for vertex cover will also be mentioned. Thi
s presentation is part of my graduate coursework on Discrete Optimization.
\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/49/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mariana Resener (SFU SEE)
DTSTART:20250206T220000Z
DTEND:20250206T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/50
DESCRIPTION:Title: Optimization in Active Distribution Grids: Modelling\, Uncertainty\,
and Applications\nby Mariana Resener (SFU SEE) as part of PIMS-CORDS
SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\
nThe challenge of solving energy distribution grid planning problems\, inc
luding operation and expansion planning\, stems from their combinatorial n
ature and the vast solution space. Several models and techniques have been
proposed in the literature to tackle these challenges. This talk explores
optimization techniques and their applications in active distribution sys
tems\, with a focus on modelling devices and distributed energy resources
(DERs)\, along with the assumptions and simplifications involved. It highl
ights the integration of these models into key optimization problems. Key
topics include a taxonomy of optimal power flow and the linearization of g
rid device models for use in mixed-integer linear programming. The present
ation also provides an overview of methods for accounting for uncertaintie
s introduced by DERs and load variations in optimization models.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/50/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jessica Stockdale (SFU)
DTSTART:20250306T220000Z
DTEND:20250306T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/51
DESCRIPTION:Title: Using Topological Features of Phylogenetic Trees to Inform Seasonal
Influenza Vaccine Design\nby Jessica Stockdale (SFU) as part of PIMS-C
ORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\nAbstra
ct: TBA\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/51/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jas Dhahan (SFU)
DTSTART:20250320T210000Z
DTEND:20250320T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/52
DESCRIPTION:Title: Simulation modelling to inform group O negative red blood cell inven
tory management in British Columbia\nby Jas Dhahan (SFU) as part of PI
MS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\n
Abstract\nBlood is a crucial life-saving product in healthcare systems. Re
d blood cells are perishable\, and managing these stocks in British Columb
ia and other regions of Canada\, with remote / rural hospitals is challeng
ing. Demand must be satisfied without wasting this resource. Group O negat
ive red blood cells are a precious resource because they can be donated un
iversally. O negative individuals comprise 6-7% of our general population\
, yet O negative demand exceeds 12% of transfusions. There is growing conc
ern over the sustainability of the O negative supply. The appropriate mana
gement of even a single red blood cell unit has the potential to save a li
fe. \n\nThere are seven health authorities in British Columbia with over 8
0 hospitals that manage their own blood inventory. British Columbia is a c
omplex jurisdiction\, which operates a provincial redistribution program\,
where red blood cells near expiry are sent from smaller to larger sites f
or use before expiring to minimize wastage.\n\nIn this talk\, we discuss o
ur ongoing collaboration with the Provincial Blood Coordination Office in
British Columbia and Canadian Blood Services to inform red blood cell inve
ntory management. We capture the key characteristics of a redistribution n
etwork of hospital blood banks using a stochastic queue network model. Our
model is calibrated to and validated against real-world data from the Tra
nsparent Blood Inventory Database. This work is funded by NSERC and the Ca
nadian Blood Services Blood Efficiency Accelerator Program.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/52/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Walaa Moursi (UBC-O hosted) (University of Waterloo)
DTSTART:20250313T210000Z
DTEND:20250313T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/53
DESCRIPTION:Title: Chambolle-Pock algorithm revisited: splitting operator and its range
with applications\nby Walaa Moursi (UBC-O hosted) (University of Wate
rloo) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture hel
d in ASB 10908.\n\nAbstract\nPrimal-dual hybrid gradient (PDHG) is a first
-order method for saddle-point problems and convex\nprogramming introduced
by Chambolle and Pock. Recently\, Applegate et al. analyzed the behavior
of\nPDHG when applied to an infeasible or unbounded instance of linear pro
gramming\, and in particular\,\nshowed that PDHG is able to diagnose these
conditions. Their analysis hinges on the notion of the\ninfimal displacem
ent vector in the closure of the range of the displacement mapping of the
splitting\noperator that encodes the PDHG algorithm. In this talk\, we dev
elop a novel formula for this range using\nmonotone operator theory. The a
nalysis is then specialized to conic programming and further to\nquadratic
programming (QP) and second-order cone programming (SOCP). A consequence
of our\nanalysis is that PDHG is able to diagnose infeasible or unbounded
instances of QP and of the ellipsoid-separation problem\, a subclass of SO
CP.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/53/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sean Kafer (remote) (Georgia Tech)
DTSTART:20250424T210000Z
DTEND:20250424T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/54
DESCRIPTION:Title: Solving 0/1 Linear Programs in Strongly Polynomial Time with Simplex
\nby Sean Kafer (remote) (Georgia Tech) as part of PIMS-CORDS SFU Oper
ations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThe Sim
plex method for solving linear programs (LPs) is one of the most widely us
ed LP solvers due to\nthe fact that it runs very quickly in practice. Howe
ver\, despite its practical efficiency and decades of study\, it remains u
nknown whether or not it provably runs in polynomial time. Other methods f
or\nsolving LPs are known to run in so-called weakly polynomial time for g
eneral LPs\, but few such results\nare known for the Simplex method even f
or very restrictive and well-known subclasses of LPs.\n\nIn this talk\, I
will discuss the special subclass of 0/1 LPs\, i.e.\, those whose vertex s
olutions have\ncomponents in {0\,1}. These are an important and widely stu
died class of LPs which model many\nproblems from combinatorial optimizati
on. Even for this important class\, the performance of the\nSimplex method
on these LPs remained unknown for decades. I will discuss the history of
their study as\nit pertains to the Simplex method\, and I will present a s
eries of results by Alex Black\, Jesús De Loera\,\nLaura Sanità\, and my
self which culminate in a proof that the Simplex method can solve 0/1 LPs
in\nstrongly polynomial time.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/54/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Anotida Madzvamuse (UBC-O hosted) (UBC)
DTSTART:20250529T210000Z
DTEND:20250529T220000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/55
DESCRIPTION:Title: Using geometric bulk-surface PDEs for insilico modelling of single a
nd collective cell migration\nby Anotida Madzvamuse (UBC-O hosted) (UB
C) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held i
n ASB 10908.\n\nAbstract\nIn this talk\, I will present insilico models fo
r 2- and 3-D cell migration\, from single to collective\, based on geometr
ic bulk-surface partial differential equations (G-BS-PDEs). The first mode
l is a geometric surface PDE approach where the cell is described by its c
ell membrane which obeys a force balance equation for its evolution. This
approach encodes naturally the biochemical processes and biomechanical pro
perties of the cells and its interactions with deformable obstacles\, and
cell-to-cell interactions. I will also present a generalisation to include
interior cell dynamics for cells migrating in confinement. The second mod
el consists of an optimal control model based on geometric multigrid metho
ds for a diffuse-interface formulation. This approach allows us to model t
he spatiotemporal dynamics of static experimental images of migrating cell
s. A by-product of this methodology is the automatic quantification of pro
liferation rates associated with cell division. A third and final approach
is a viscoelastic model\, where the displacements of the cell are driven
by biomolecular species which obey a reaction-diffusion system. Numerical
results will be presented to illustrate the novelty of these mechanobioche
mical models for single and collective cell migration. Single and collecti
ve cell migration are essential for physiological\, pathological and biome
dical processes in development\, repair\, and disease\, for example\, in e
mbryogenesis\, wound healing\, immune response\, cancer metastasis\,\ntumo
ur invasion\, and inflammation.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/55/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Clement Royer (UBC-O hosted\, on-line only) (Université Paris Dau
phine-PSL)
DTSTART:20250729T180000Z
DTEND:20250729T190000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/56
DESCRIPTION:Title: A derivative-free method for continuous submodular optimization\
nby Clement Royer (UBC-O hosted\, on-line only) (Université Paris Dauphin
e-PSL) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture he
ld in ASB 10908.\n\nAbstract\nSubmodular functions are a classical concept
of discrete optimization\, that can also be extended to the\ncontinuous s
etting. In particular\, the class of continuous submodular functions encom
passes some\nnonconvex functions arising in natural language processing\,
which partly explains renewed interest for\nthis topic in recent years.\n\
nIn this talk\, I will describe a derivative-free algorithm for continuous
submodular optimization\nover compact sets\, adapted from a classical fra
mework for bound-constrained derivative-free\noptimization. The first part
will focus on theoretical (complexity) guarantees for the proposed method
\,\nwhich departs from the general nonconvex setting. The second part will
illustrate the practical\nperformance of our algorithm on continuous subm
odular tasks. Time permitting\, I will also discuss the\ndiscrete submodul
ar case.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/56/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ali Hassanzadeh (jointly hosted with Beedie's Technology and Opera
tions Management Area) (University of Manchester)
DTSTART:20250815T170000Z
DTEND:20250815T190000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/57
DESCRIPTION:Title: From Fixtures to Fairness: Analytics-Driven Decision Making in Profe
ssional Sports\nby Ali Hassanzadeh (jointly hosted with Beedie's Techn
ology and Operations Management Area) (University of Manchester) as part o
f PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.
\n\nAbstract\nNote the seminar meets at an unusual location\, WMC 4335<
/B>.\n\nTitle: From Fixtures to Fairness: Analytics-Driven Decision Mak
ing in Professional Sports\n\nProblem definition: Professional sports
leagues may be suspended because of various reasons\, such as the recent c
oronavirus disease 2019 pandemic. A critical question that the league must
address when reopening is how to appropriately select a subset of the rem
aining games to conclude the season in a shortened time frame. Despite the
rich literature on scheduling an entire season starting from a blank slat
e\, concluding an existing season is quite different. Our approach attempt
s to achieve team rankings similar to those that would have resulted had t
he season been played out in full. Methodology/results: We propose a data-
driven model that exploits predictive and prescriptive analytics to produc
e a schedule for the remainder of the season composed of a subset of origi
nally scheduled games. Our model introduces novel rankings-based objective
s within a stochastic optimization model\, whose parameters are first esti
mated using a predictive model. We introduce a deterministic equivalent re
formulation along with a tailored Frank–Wolfe algorithm to efficiently s
olve our problem as well as a robust counterpart based on min-max regret.
We present simulation-based numerical experiments from previous National B
asketball Association seasons 2004–2019\, and we show that our models ar
e computationally efficient\, outperform a greedy benchmark that approxima
tes a non-rankings-based scheduling policy\, and produce interpretable res
ults. Managerial implications: Our data-driven decision-making framework m
ay be used to produce a shortened season with 25%–50% fewer games while
still producing an end-of-season ranking similar to that of the full seaso
n\, had it been played.\n\nLink to paper: https://pubsonline.informs.org/d
oi/abs/10.1287/msom.2022.0558\n\n \n\nPart II: NBA Expansion: Opportuni
ties to Reform the League\n\nIn this study\, we explore how the NBA co
uld restructure its divisions and conferences in light of potential league
expansion. Building on optimization models\, we consider two fairness-ori
ented formulations: a total travel distance minimization and a Nash bargai
ning framework that balances travel burden across teams\, as well as distr
ibution of media market size. Our approach evaluates realignment scenarios
using geographic clustering and provides insights into how fairness and e
fficiency can be reconciled in league design. This work highlights the val
ue of data-driven approaches in making strategic structural decisions for
professional sports leagues.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/57/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yuan Zhou (remote) (University of Kentucky)
DTSTART:20251202T233000Z
DTEND:20251203T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/58
DESCRIPTION:Title: All Cyclic Group Facets Inject\nby Yuan Zhou (remote) (Universit
y of Kentucky) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLe
cture held in ASB 10908.\n\nAbstract\nIn this talk\, we study cut-generati
ng functions in the setting of the Gomory-Johnson group relaxations\nfor i
nteger programming. We address an open question: whether every facet (extr
eme function) for a\nfinite cyclic group relaxation injects into the space
of extreme functions for the infinite group problem. We give a variant o
f the Basu-Hildebrand-Molinaro approximation theorem [IPCO 2016] for\ncont
inuous minimal functions of the infinite group problem. Specifically\, we
show that any piecewise\nlinear minimal function with rational breakpoints
in 1/qZ and rational values at these breakpoints can be approximated by p
iecewise linear two-slope extreme functions while preserving all function
values on\n1/qZ: a feature not guaranteed by the earlier construction. As
a corollary\, every extreme function for the finite group problem on 1/qZ
is the restriction of a continuous piecewise linear two-slope extreme\nfun
ction for the infinite group problem\, with breakpoints on a refinement 1/
(Mq)Z. Combined with\nGomory’s master theorem\, this establishes that th
e infinite group problem indeed serves as the correct\nmaster problem for
facets of one-row group relaxations.\n\nThis is a joint work with Matthias
Koeppe.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/58/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Krisztina Vásárhelyi (Vancouver Coastal Health)
DTSTART:20251007T223000Z
DTEND:20251007T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/59
DESCRIPTION:Title: Embedded Research Program in Healthcare Operations Research: An SFU/
VCH Partnership\nby Krisztina Vásárhelyi (Vancouver Coastal Health)
as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in A
SB 10908.\n\nAbstract\nThe SFU Operations Research Clinic capstone project
course instructors have been in partnership with Vancouver Coastal Health
(VCH) since 2018. Each Spring term a cohort of students works on a proble
m posed by the VCH health authority partners. This successful collaboratio
n has grown into a broader partnership that fosters embedded operations re
search (OR) at VCH. The Embedded Research Program is one key initiative th
at brings two SFU faculty and two graduate students to VCH\, where they wo
rk closely with health system partners on priority operational problems. T
his structured program and the Operations Research Clinic research project
s have generated actionable findings that have informed capacity managemen
t and operations at VCH. This experience brought many valuable lessons abo
ut the barriers and facilitators of embedding research into health care pr
actice and I will share some key insights in this talk. With the integrati
on of this understanding of what works and what doesn’t\, there is poten
tial for further expansion of this SFU/VCH partnership to support health c
are operations at VCH.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/59/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Peter Zhang (UBC-O hosted) (Carnegie Mellon University)
DTSTART:20251118T233000Z
DTEND:20251119T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/60
DESCRIPTION:Title: Robust Paths: Geometry and Computation\nby Peter Zhang (UBC-O ho
sted) (Carnegie Mellon University) as part of PIMS-CORDS SFU Operations Re
search Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nApplying robust
optimization often requires selecting an appropriate uncertainty set both
in shape and size\, a choice that directly affects the trade-off between a
verage-case and worst-case performances. In practice\, this calibration is
usually done via trial-and-error: solving the robust optimization problem
many times with different uncertainty set shapes and sizes\, and examinin
g their performance trade-off. This process is computationally expensive a
nd ad hoc. In this work\, we take a principled approach to study this issu
e for robust optimization problems with linear objective functions\, conve
x feasible regions\, and convex uncertainty sets. We introduce and study w
hat we define as the robust path: a set of robust solutions obtained by va
rying the uncertainty set's parameters. Our central geometric insight is t
hat a robust path can be characterized as a Bregman projection of a curve
(whose geometry is defined by the uncertainty set) onto the feasible regio
n. This leads to a surprising discovery that the robust path can be approx
imated via the trajectories of standard optimization algorithms\, such as
the proximal point method\, of the deterministic counterpart problem. We g
ive a sharp approximation error bound and show it depends on the geometry
of the feasible region and the uncertainty set. We also illustrate two spe
cial cases where the approximation error is zero: the feasible region is p
olyhedrally monotone (e.g.\, a simplex feasible region under an ellipsoida
l uncertainty set)\, or the feasible region and the uncertainty set follow
a dual relationship. We demonstrate the practical impact of this approach
in two settings: portfolio optimization and adversarial deep learning.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/60/
END:VEVENT
BEGIN:VEVENT
SUMMARY:14 speakers
DTSTART:20251018T155000Z
DTEND:20251018T230000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/61
DESCRIPTION:Title: WCOM 2025 at UBC-Okanagan\nby 14 speakers as part of PIMS-CORDS
SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\
nFor details click here<
/a>.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/61/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Daniela Lubke (remote) (University of Waterloo)
DTSTART:20251028T223000Z
DTEND:20251028T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/62
DESCRIPTION:Title: Integrating Machine Scheduling and Personnel Allocation in a Large-S
cale Analytical Services Facility via Column Generation\nby Daniela Lu
bke (remote) (University of Waterloo) as part of PIMS-CORDS SFU Operations
Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThis work inv
estigates the integration of machine scheduling and personnel allocation p
roblems. In\nmachine scheduling\, the goal is to find the optimal assignme
nt of jobs to machines within a given time\nhorizon\, considering processi
ng times and machine capacities in a time-discretized model. Personnel\nal
location problems aim to determine the best employee distribution while re
specting business\,\nregulatory\, or satisfaction constraints (for example
\, maximum work hours for an employee in a specific\nactivity\, the durati
on an employee can remain in one activity without rest\, or preferred work
days). These two problems are inherently connected and should ideally be a
ddressed together in a unified\nformulation\, though this is computational
ly demanding. This work introduces a column generation-based\nalgorithm de
signed to find good quality solutions for an industrial-scale analytical s
ervices facility\nwhere integrated machine scheduling and personnel alloca
tion decisions are made daily. Our\ncomputational results indicate that th
e proposed approach consistently produces high-quality solutions\neven as
the problem size increases\, outperforming other existing methods.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/62/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nicholas Richardson (UBC)
DTSTART:20260120T233000Z
DTEND:20260121T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/63
DESCRIPTION:Title: Optimization and applications for unsupervised signal demixing\n
by Nicholas Richardson (UBC) as part of PIMS-CORDS SFU Operations Research
Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nThroughout scientific
and commercial domains\, we are often interested in separating mixed signa
ls into their component sources. Supervised deep learning is state-of-the-
art when large and well-labeled datasets can be used. But in many applicat
ions\, large scale collection and labelling can be too impraticable\, expe
nsive\, or behind copyright laws. This talk will explore a number of appli
cations from sediment analysis\, genome sequencing\, and audio source sepa
ration that fall into the scarce data category. We will see a few approach
es I have used to model and solve these problems such as sparse feature mo
dels and tensor factorizations. These unsupervised learning techniques avo
id a training phase and have the advantage of adapting to the specific exa
mple at hand.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/63/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Pengcheng Xie (remote) (Lawrence Berkeley National Laboratory)
DTSTART:20260127T233000Z
DTEND:20260128T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/64
DESCRIPTION:Title: Model-Based Derivative-Free Optimization with Improved Approximation
\nby Pengcheng Xie (remote) (Lawrence Berkeley National Laboratory) as
part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB
10908.\n\nAbstract\nThis talk will discuss traditional and modern approxi
mation techniques for (expensive) derivative-free\noptimization\, in which
the approximation serves as a tool for identifying the optimality of the
black-box objective\, rather than merely approximating the black box itsel
f. Static (offline) strategies and dynamic (online) strategies will be dis
cussed. Such approximation ideas can also be naturally extended to first-o
rder and higher-order methods.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/64/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yuriy Zinchenko (Gurobi and University of Calgary)
DTSTART:20251125T233000Z
DTEND:20251126T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/65
DESCRIPTION:Title: Can an infeasible MIP solve itself?\nby Yuriy Zinchenko (Gurobi
and University of Calgary) as part of PIMS-CORDS SFU Operations Research S
eminar\n\nLecture held in ASB 10908.\n\nAbstract\nThe analysis of why a sp
ecific MIP instance is infeasible formally can be reduced to computing an
Irreducible Infeasible Subset (IIS) of the constraints. Unlike the case of
LP\, for MIP there is no useful duality that can be employed to facilitat
e such computations. The process of determining an IIS for MIP is typicall
y handled with brute force\, e.g.\, by use of deletion filters and alike\,
thus rendering IIS determination for a MIP into a much harder computation
al task. We will discuss one approach to optimizing this process and what
components of this approach could make it into the newest version of Gurob
i.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/65/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mohammad Delasay (room: WMC 5302) (Stony Brook University and Univ
ersity of Alberta)
DTSTART:20251121T200000Z
DTEND:20251121T213000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/66
DESCRIPTION:Title: Delay Information Sharing in Two-sided Platforms\nby Mohammad De
lasay (room: WMC 5302) (Stony Brook University and University of Alberta)
as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in A
SB 10908.\n\nAbstract\nWe analyze how strategic customers and providers r
espond to delay information in a matching system. Using Markovian queueing
models and equilibrium analysis\, we evaluate three disclosure policies (
no information\, binary\, and occupancy) and identify conditions under whi
ch each improves match rates. The results also reveal misalignments betwee
n platform-optimal choices and user preferences.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/66/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Youssef Diouane (remote) (Polytechnique Montréal)
DTSTART:20260224T233000Z
DTEND:20260225T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/67
DESCRIPTION:Title: Direct-Search for Min-Max Derivative-Free Optimization\nby Youss
ef Diouane (remote) (Polytechnique Montréal) as part of PIMS-CORDS SFU Op
erations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nRecen
t applications in machine learning have renewed the community’s interest
in min-max\noptimization problems. While gradient-based optimization meth
ods are widely used to solve these\nproblems\, there exist many scenarios
where such techniques are not well suited\, or even not applicable\,\npart
icularly when gradients are not accessible. In this talk\, we will investi
gate the use of direct-search methods\, which belong to a class of derivat
ive-free techniques that only require access to the objective function thr
ough an oracle. We will present a novel direct-search method for min-max s
addle-point problems\, where the min and max players are updated sequentia
lly. The convergence of this algorithm will be discussed in both determini
stic and stochastic settings. Finally\, experimental results related to ro
bust optimization and Generative Adversarial Networks will be presented to
illustrate how the proposed method can outperform commonly used optimizat
ion schemes.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/67/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Amzi Jeffs (remote) (Pacific Northwest National Laboratory)
DTSTART:20260324T223000Z
DTEND:20260324T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/68
DESCRIPTION:Title: AI-Based Algorithms for Real-Time Anomaly Detection in Ground Penetr
ating Radar Data\nby Amzi Jeffs (remote) (Pacific Northwest National L
aboratory) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLectur
e held in ASB 10908.\n\nAbstract\nGeophysical inversion is a well-establis
hed method for conducting non-invasive surveys of underground\nanomalies\,
such as buried infrastructure. Typically\, inversion is performed using n
umerical methods:\ngradient descent and physics-based simulations are used
to create a subsurface model that accurately\nexplains the surface survey
s. However\, these methods are extremely computationally expensive\,\nsome
times taking days or weeks to complete. Additionally\, the inverse problem
s are highly\nunderdetermined\, and must be carefully regularized based on
expert input. This creates a challenging\nbottleneck for field operators\
, who must accept lengthy delays between conducting surveys and starting\n
work.\n\nAI-based methods can help overcome these challenges. We will disc
uss advances in this direction based\non work for the Grid Overhaul with P
roactive\, High-Speed Undergrounding for Reliability\, Resilience\,\nand S
ecurity (GOPHURRS) project. Our approach uses thousands of geophysical fie
ld surveys to train\nconvolutional neural networks that can perform subsur
face anomaly detection in near real-time. We will\ndiscuss the strengths a
nd limitations of this AI-based approach\, best practices for using physic
al data\nwith AI\, and our approach to optimizing our overall workflow.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/68/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Fredrik Odegaard (room: WMC 4335) (Ivey Business School (Universit
y of Western Ontario))
DTSTART:20260220T193000Z
DTEND:20260220T210000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/69
DESCRIPTION:Title: Decoding Consumer Preferences Using Attention-Based Language Models<
/a>\nby Fredrik Odegaard (room: WMC 4335) (Ivey Business School (Universit
y of Western Ontario)) as part of PIMS-CORDS SFU Operations Research Semin
ar\n\nLecture held in ASB 10908.\n\nAbstract\nThis paper proposes a new de
mand estimation method using attention-based language models. An encoder-o
nly language model is trained in a two-stage process to analyze the natura
l language descriptions of used cars from a large US-based online auction
marketplace. The approach enables semi-nonparametrically estimation for th
e demand primitives of a structural model representing the private valuati
ons and market size for each vehicle listing. In the first stage\, the lan
guage model is fine-tuned to encode the target auction outcomes using the
natural language vehicle descriptions. In the second stage\, the trained l
anguage model's encodings are projected into the parameter space of the st
ructural model. The model's capability to conduct counterfactual analyses
within the trained market space is validated using a subsample of withheld
auction data\, which includes a set of unique "zero shot" instances.\n\nT
his is joint work with Joshua Foster.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/69/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ahmet Alacaoglu (UBC)
DTSTART:20260303T233000Z
DTEND:20260304T003000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/70
DESCRIPTION:Title: Towards Weaker Variance Assumptions for Stochastic Optimization: A B
last From the Past\nby Ahmet Alacaoglu (UBC) as part of PIMS-CORDS SFU
Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nIn
this talk\, I will present some recent advances for analyzing stochastic
optimization methods without the bounded variance assumption. It is well-k
nown that the bounded variance assumption is violated for even the most st
andard problems such as linear least squares problem. We will see that the
analysis for obtaining optimal rates of convergence under realistic varia
nce assumptions builds on a connection between the classical literatures f
or stochastic approximation and the Halpern iteration for solving fixed-po
int problems. We will discuss the extensions to proximal algorithms for so
lving regularized problems and stochastic convex nonlinear programs\, as w
ell as the required ideas for getting rate guarantees on the last iterate
of the algorithm\, which is widely used in practice.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/70/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jeremy Chiu (SFU)
DTSTART:20260317T223000Z
DTEND:20260317T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/71
DESCRIPTION:Title: The Burden of Tuberculosis among Foreign-born Canadians - Estimates
with Dynamic Models\nby Jeremy Chiu (SFU) as part of PIMS-CORDS SFU Op
erations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nDespi
te only comprising about a quarter of the total population of Canada\, for
eign-born individuals bear about three-quarters of the burden of active tu
berculosis (TB) cases. To investigate the impact of immigration on the bur
den of TB among foreign-born Canadians\, we develop an SEIR-compartment mo
del that distinguishes between actively infected\, latently infected\, and
uninfected individuals. Unknown parameters are calibrated to reports on
the incidence and prevalence of active TB in Canada. We validate our mo
del by comparing model computed quantities to other estimates of tuberculo
sis burden among foreign-born Canadians\, including an estimate of the pre
valence of latent TB infection (LTBI) among immigrants entering Canada.
Our model predicts that among the foreign-born population\, Canada will no
t meet the End TB 2035 goal of reducing incidence by 90% compared to 2015\
, primarily due to the activation of foreign-born Canadians with LTBI.\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/71/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Hongda Li (UBC-O hosted) (UBC Okanagan)
DTSTART:20260407T223000Z
DTEND:20260407T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/72
DESCRIPTION:Title: Quadratic Growth Gives Near Optimal Total Complexity for Inexact Acc
elerated Proximal Gradient Method\nby Hongda Li (UBC-O hosted) (UBC Ok
anagan) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture h
eld in ASB 10908.\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/72/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Temesgen Abraha and Tanmaya Karmarkar (UBC-O hosted) (UBC Okanagan
)
DTSTART:20260414T223000Z
DTEND:20260414T233000Z
DTSTAMP:20260404T110831Z
UID:SFUOR/73
DESCRIPTION:by Temesgen Abraha and Tanmaya Karmarkar (UBC-O hosted) (UBC O
kanagan) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture
held in ASB 10908.\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/SFUOR/73/
END:VEVENT
END:VCALENDAR