2016 INFORMS Annual Meeting Program

SC21

INFORMS Nashville – 2016

SC20 106C-MCC

2 - Generating And Solving The Large Scale AC-security Constrained Optimal Power Flow Problems In Parallel Feng Qiang, Argonne National Laboratory, Lemont, IL, 60439,

United States, fqiang@anl.gov, Cosmin G Petra, Joseph A Huchette, Miles Lubin, Mihai Anitescu

Novel Dimension Reduction Techniques for High Dimensional Data Using Information Complexity

In this talk, we present an integrated approach for the modelling and solution of the AC-SCOPF problems using StructJuMP, a newly developed parallel extension of JuMP for modelling large scale optimization problems in Julia and PIPS optimization solver for HPC platforms. We will present a thorough study of the the parallel performance of StructJuMP and PIPS for large scale AC-SCOPF instances on hundreds of nodes. 3 - Paraxpress: A Massively Parallelized MIP Solver Designed To Run On The Largest Supercomputers Yuji Shinano, Zuse Institute Berlin, Takustrasse 7, Berlin, 14195, Germany, shinano@zib.de, Timo Berthold, Stefan Heinz The Ubiquity Generator (UG) is a framework for the external parallelization of MIP solvers. It was used to develop ParaSCIP, a distributed memory, massively parallel version of the open source solver SCIP, that runs on up to 80,000 cores in parallel. In this talk, we introduce ParaXpress, for which one of the fastest commercial MIP solvers, the FICO Xpress-Optimizer, has been parallelized by UG. Combining the internal shared-memory parallelization of Xpress and the external parallelization of UG, we aim at a new order of magnitude for supercomputer core-usage in MIP solving. SC19 106B-MCC Computation and Theory in Network Optimization and Analysis Sponsored: Computing Sponsored Session Chair: Cole Smith, Clemson University, Clemson University, Clemson, SC, 29634, United States, jcsmith@clemson.edu 1 - Models And Algorithms For Maximum Proportional Flow Problems With Semicontinuous Restrictions We consider a variation of the multi-source, multi-sink maximum flow problem in which flow must emanate from the source nodes according to a prescribed rate, while flow arrives to the sink nodes at another given rate. Additionally, we restrict flow variables to be semicontinuous, in which the flow must either be 0 or no less than some lower bound. We call this problem the semicontinuous maximum proportional flow problem (SC-MPFP) since the amount of outgoing flow must leave the source nodes and arrive at the sink nodes according to a given proportional pattern. To solve the SC-MPFP, we decompose the formulation and employ a Branch-and-Price algorithm. 2 - Enumeration Algorithms For Infrastructure Resilience Analysis W Matthew Carlyle, Naval Postgraduate School, mcarlyle@nps.edu, David Alderson We propose a functional definition of infrastructure resilience based on parametric analysis of two-stage (attacker-defender) and three-stage (defender- attacker-defender) models that require enumeration of a potentially enormous number of optimization problems. We present computational techniques that use bounding arguments to significantly limit the enumeration while still providing useful measures of infrastructure resilience and support the use of faster heuristic algorithms for the most difficult of these problems. 3 - Faster Algorithms For The Time-cost-tradeoff Problem And Minimum Cost K-flow Problems With A New All-min-Cuts Procedure Dorit S. Hochbaum, University of California, Berkeley, Berkeley, CA, United States, hochbaum@ieor.berkeley.edu We explore surprising links between the time-cost-tradeoff (TCT) problem in project management and the minimum cost flow problem (MCF) leading to faster algorithms for both problems. The algorithm relies on a new procedure all-min- cuts procedure, which for a given maximum flow, is capable of generating all minimum cuts of equal value very efficiently. This results in faster strongly polynomial algorithms for unit capacity MCF, the K-MCF problem and uniform costs TCT and match the complexity of the fastest algorithm for the assignment problem. Robert Mark Curry, Clemson University, Clemson, SC, United States, rmcurry@g.clemson.edu, Cole Smith

Invited: Tutorial Invited Session

Chair: Hamparsum Bozdogan, University of Tenneesse-Knoxville, Oper and Mgmt Sci, Knoxville, TN, 37996, United States, bozdogan@utk.edu 1 - Novel Dimension Reduction Techniques For High Dimensional Data Using Information Complexity

Hamparsum Bozdogan, University of Tennessee-Knoxville, Oper and Mgmt Sci, Knoxville, TN, 37996, United States, bozdogan@utk.edu, Esra Pamukcu

This tutorial introduces and develops two computationally feasible intelligent feature extraction techniques that addresses the potentially daunting statistical and combinatorial problems. First part of the tutorial employs a three-way hybrid between: Probabilistic Principal Component Analysis (PPCA) to reduce the dimensionality of the dependent variables; Multivariate regression (MVR) models that account for misspecification of the distributional assumption to determine a predictive operating model for glass composition for automobiles; and uses the genetic algorithm (GA) as the optimizer along with the misspecification-resistant form of Bozdogan’s ICOMP as the fitness function. Second part of the tutorial is devoted to dimension reduction via a novel Adaptive Elastic Net (AEN) regression model to reduce the dimension of a Japanese stock index called TOPIX as the response to build a best predictive model when we have “large p, small n” problem. Our results show the remarkable dimension reduction in both of these real-life examples of wide datasets, which demonstrates the versatility and the utility of the two proposed novel statistical data modeling techniques. SC21 107A-MCC Monitoring and Prevention of Healthcare Associated Infections Sponsored: Health Applications Sponsored Session Chair: Eduardo Perez, Texas State University, 1, San Marcos, TX, 1, United States, eduardopr@txstate.edu 1 - Optimal Pooling Strategies For Nucleic Acid Testing Of Donated Blood Considering Viral Load Growth Curves And Donor Characteristics Hadi El-Amine, Virginia Tech, Blacksburg, VA, United States, hadi@vt.edu, Hadi El-Amine, George Mason University, Fairfax, VA, 22030, United States, hadi@vt.edu, Ebru Korular Bish, Douglas R Bish Blood product safety, in terms of being free of transfusion-transmittable infections (TTIs), is crucial. Nucleic Acid Testing (NAT) technology enables earlier detection of infections but is more expensive, hence, most blood centers administer NAT to pools of blood samples from multiple donors. Since some donor characteristics are uncertain, we develop a chance-constrained model that determines the optimal NAT pool sizes for various TTIs, considering both non-universal (where first-time donors undergo more extensive screening), and universal (i.e., common testing for all donors) strategies, so as to minimize the TTI risk, while remaining within the testing budget with a high probability. 2 - Infection Control In Outpatient Clinics: The Risk- Efficiency Tradeoff Cory Stasko, Massachusetts Institute of Technology, cstasko@mit.edu, As HAIs remain a major problem and U.S. healthcare shifts towards outpatient, understanding the risk of HAIs in this environment is important. For a particular clinic, we simulate the potential impacts of two interventions: 1) improving hand hygiene, and 2) separating likely infectious patients from other patients. By creating an integrated discrete event and agent based model to simulate patient flow and the spread of infections, we examine how these two domains interact, testing 891 intervention combinations in terms of wait time, infection exposures, and implementability. Interdependent effects and the tradeoff between risk reduction and operational efficiency are discussed.

75