Date	Speaker	Title and Abstract
August 23rd *Friday* *4:00 p.m.*	Li Xia Sun Yat-Sen University	The Mean-Variance Optimization of Markov Decision Processes Abstract: In this talk, we study the optimization of an infinite stage discrete time Markov decision process (MDP) with a long-run average metric considering both mean and variance of rewards together. Such performance metric is important since the mean indicates average returns and the variance indicates risk or fairness. However, the variance metric couples the rewards at all stages, this optimization problem is not a standard MDP and the traditional dynamic programming fails. We study this problem from the viewpoint of sensitivity-based optimization. We derive a performance difference formula which quantifies the difference of the mean-variance combined metrics under any two different policies. The correlation in the variance of rewards at different stages can be decoupled through a square term in the difference formula. A necessary condition of the optimal policy is derived. With the difference formula, we develop a policy iteration type algorithm to iteratively improve the combined performance metric. We prove that this algorithm can converge to the local optimum in randomized policy space. Furthermore, we introduce exploration mechanisms to make the algorithm be able to jump from local optima to global optimum. Efficient frontiers of Pareto optimization can also be derived, which enhances the understanding to this two-objective optimization problem. Finally, we apply our approach to study the fluctuation reduction of output power of renewables and energy storage systems, which demonstrates the main idea of this paper.
August 2nd *Friday* *1:30* *Lib 2020* (Burnaby)	Michelle Spencer M.Sc. thesis defence Senior Supervisor: T. Stephen	Formulating Quadratic Traveling Salesman Problems for Computation Abstract: The Traveling Salesman Problem (TSP) is a fundamental combinatorial optimization problem. Adding costs associated with pairs of edges included in a tour gives the Quadratic Traveling Salesman Problem (QTSP). This increases modeling power by allowing, for example, the inclusion of transfer costs between edges. We consider a general version of this problem, where costs are attached to all pairs of edges. We give a brief history of computational solvers, especially in relation to the TSP. For the QTSP, we consider modifying the structure of the quadratic cost input and linearizing the quadratic objective function, with detail on how to generate the modifications and linearizations. We study the impact of these structures on computational efficiency for randomly generated instances, using the Gurobi solver. We find that by making the quadratic cost matrix negative semidefinite, we improve solve times, and that solving the problem as a quadratic minimization problem outperforms linearization approaches.
August 1st *11:00 a.m.*	Zsolt Saffer Institute of Statistics and Mathematical Methods in Economics TU Wein	Probabilistic capacity control of queues Abstract: This talk is about controlling the capacity of a queueing system. Capacity of a queueing system can be modeled on different ways, like number of servers or service rate. We review several different ways of controlling such capacity, like making the capacity or the change of capacity dependent on the system state, like e.g. the number of customers in the system. In this talk we concentrate rather on the second way, in which the capacity is allowed to be increased and decreased by a fixed value at each customer service completion. We present the analysis of two queueing models with such controllable capacity. In the first model, the number of active servers can be controlled by means of probabilities specifying the dependency of the number of active servers on the actual number of customers and the number of active servers. The service time is constant and the concurrently served customers are served in synchronized manner. The active number of servers can be incremented, decremented or kept unchanged at the ends of service time according to the given probabilities. The system is a loss system, i.e. it has no buffer for long-term customer waiting. We provide explicit form results for the joint and marginal distributions of the number of servers and the number of customers on PGF level as well as expressions of the most important system measures. The second queueing analysis considers an M/M/1 queue, in which the customer service rate is allowed to be increased and decreased by a fixed value at each customer service completion. These changes in service rate are controlled by probabilities depending on the actual number of customers and the actual service rate. We establish a methodology which utilizes the specific structure of the model. This methodology inherits some element from the stationary analysis of the standard QBD model and provides a first order, forward algorithm for computing the stationary probability vectors of the number of customers in the system. We derive also the vector probability generating function and the vector mean of the stationary number of customers. Such queueing models with controllable capacity can be applied e.g. in intelligent transportation systems or manufacturing systems, in which the processing speed follows the processing demand.
Tuesday, June 4th Joint O.R. and Discrete Math Seminar *Burnaby* *1:30* *AQ K9509*	Ahmad Abdi Tepper School of Business Carnegie Mellon University and Department of Mathematics London School of Economics	Ideal clutters and k-wise intersecting families Abstract: A clutter is *ideal* if the corresponding set covering polyhedron has no fractional vertices, and it is *k-wise intersecting* if the members don’t have a common element but every k members do. We conjecture that there is a constant k such that every k-wise intersecting clutter is non-ideal. I will show how this conjecture for k=4 would be an extension of Jaeger’s 8-flow theorem, and how a variation of the conjecture for k=3 would be an extension of Tutte’s 4-flow conjecture. I will also discuss connections to tangential 2-blocks, binary projective geometries, the sums of circuits property, etc.
April 25th	Jean-Philippe Richard Industrial and Systems Engineering University of Minnesota	On the Convexification of Permutation-invariant Sets arising in MINLP and Data Problems Abstract: Permutation-invariant sets - sets that do not change when the variables are permuted - appear in a variety of optimization problems, including sparse principal component analysis. Solving permutation-invariant optimization problems to global optimality requires the derivation of strong convex relaxations for their feasible region. In this talk, we study how to construct the convex hull of permutation-invariant sets in a higher dimensional space. When projection is possible, we also obtain convex hull descriptions in the original space of variables. We illustrate our techniques by developing (1) a novel reformulation and relaxation for sparse principal component analysis, (2) convex hull of matrices with bounded rank and spectral norms, (3) convex envelopes of multi-linear functions over certain domains, and (4) linear descriptions of logical and sparsity constraints arising in the formulation of 0-1 mixed integer programs. This is joint work with Mohit Tawarmalani (Purdue) and Jinhak Kim (University of South Alabama).
April 5th *Friday, 1:30* *RCB 6125* (Burnaby)	Liam Erdos, Ben Gregson, Shane Jace and Martin Zhu Simon Fraser University	Math 402W Operations Research Clinic project presentation Coordinating Primary Care Operating Hours to Reduce Acute Care Visits
February 28th *Thursday* *9:30-12:00* *SUR 3040*	Xiaorui Li Ph.D. thesis defence Senior Supervisor: Z. Lu	Sparse and Low Rank Approximation via Partial Regularization: Models, Theory and Algorithms Abstract: Sparse representation and low-rank approximation are fundamental tools in fields of signal processing and pattern analysis. In this thesis, we consider introducing some partial regularizers to these problems in order to neutralize the bias incurred by some large entries (in magnitude) of the associated vector or some large singular values of the associated matrix. In particular, we first consider a class of constrained optimization problems whose constraints involve a cardinality or rank constraint. Under some suitable assumptions, we show that the penalty formulation based on a partial regularization is an exact reformulation of the original problem in the sense that they both share the same global minimizers. We also show that a local minimizer of the original problem is that of the penalty reformulation. Specifically, we propose a class of models with partial regularization for recovering a sparse solution of a linear system. We then study some theoretical properties of these models including existence of optimal solutions, sparsity inducing, local or global recovery and stable recovery. In addition, numerical algorithms are proposed for solving those models, in which each subproblem is solved by a nonmonotone proximal gradient (NPG) method. Despite the complication of the partial regularizers, we show that each proximal subproblem in NPG can be solved as a certain number of one-dimensional optimization problems, which usually have a closed-form solution. The global convergence of these methods are also established. Finally, we compare the performance of our approach with some existing approaches on both randomly generated and real-life instances, and report some promising computational results.
PIMS - SFU Seminar December 6th *SUR 5380*	Asia Ivic Weiss York University	Regular Polyhedra, Polytopes and Beyond Abstract: In this talk we summarize the classification of regular polyhedra and polytopes and extend the concept to that of hypertope: a thin, residually connected incidence geometry. We present the characterization of the automorphism groups of regular hypertopes and overview recent results on classification of toroidal hypertopes.
October 6th *Saturday, 8:30-4:00* *Kelowna*	WCOM Hosted by UBC Okanagan	Details of the Fall 2018 West Coast Optimization Meeting available from the hosts.
PIMS - SFU Seminar October 4th	Jong-Shi Pang Daniel J. Epstein Department of Industrial and Systems Engineering University of Southern California	Non-problems in Optimization for Statistics Abstract: The phrase "non-problems" in the title refers to a collection of adjectives that start with the prefix "non". These include "non-convex", "non-differentiable", "non-traditional", "non-trivial", and "non-stochastic gradient" (as a counter to a topical research theme), all in the context of optimization for statistics. Outside a stand-alone optimization problem, the phrase could include "non-cooperative" game theory as this is an area where there are significant research opportunities when combined with the other "non-subjects". I will present a variety of these non-problems and give a brief summary of my research on them.