2015 Informs Annual Meeting

MA16

INFORMS Philadelphia – 2015

MA16 16-Franklin 6, Marriott

3 - Network Optimization under Non-linear Utility Chin Hon Tan, National University of Singapore, 1 Engineering Drive, Singapore, 117576, Singapore, isetch@nus.edu.sg Decision makers are rarely risk neutral in practice. Hence, solutions that maximize expected rewards or minimize expected costs, which assumes that the decision maker is risk neutral, may not be appropriate. In this presentation, we study the sensitivity of solutions with respect to the risk neutral assumption and discuss how solutions that are robust to the decision maker’s utility can be improved upon. 4 - An Integer Programming Approach for Mixed Fault Diameters Elham Sadeghi, Graduate Research Assistant, Univesity of Arizona, Tucson, AZ, United States of America, sadeghi@email.arizona.edu, Neng Fan We consider the minimum (k,l)-connected d-dominating set problem which is a fault-tolerance dominating set. This problem is a generalization of minimum connected dominating set problem. The integer programming formulations based on vertex-cut and edge-cut is introduced and a cutting plane algorithm is proposed to solve it. MA18 18-Franklin 8, Marriott The Reborn of Traditional OR Methods in the Era of Big Data Cluster: Modeling and Methodologies in Big Data Invited Session Chair: Shouyi Wang, Assistant Professor, University of Texas at Arlington, 3105 Birch Ave, Grapevine, TX, 76051, United States of America, shouyiw@uta.edu Chair: Danica Xiao, PhD Candidate, University of Washington, Seattle, 3900 Northeast Stevens Way, Seattle, WA, 98195, United States of America, xiaoc@uw.edu 1 - A Dynamic Active-Set Method for Linear Programming Alireza Noroziroshan, University of Texas at Arlington, 600 Grand Ave, Apt#103, Arlington, TX, 76010, United States of America, alireza.norozi.en@gmail.com, Bill Corley, Jay Rosenberger An active-set method obtains solution for linear programming problems by adding one or more constraints at a time to solve smaller problems iteratively. We present an e?cient constraint selection rule for adding varying numbers of constraints at each iteration. This approach is significantly faster than the standard linear programming algorithms. 2 - Big Data Analytics for RFID-Enabled Logistics Data from Ubiquitous Manufacturing Shopfloors Ray Y. Zhong, Post-doctoral Fellow, The University of Hong Kong, 8-16 Haking Wong Building, IMSE, Pokfulam Road, HKU, Hong Kong, Hong Kong - PRC, zhongzry@gmail.com, George Q. Huang RFID has been widely used in logistics and supply chain management. This paper discusses the manufacturing shopfloor where typical logistics resources are converted into smart manufacturing objects (SMOs) by using RFID and wireless technologies to create a RFID-enabled intelligent shopfloor environment. In such environment, enormous RFID data has been captured and collected. This paper introduces a Big Data Analytics for the RFID logistics data by defining different behaviors of the SMOs. 3 - On the Mixed Set Covering, Packing and Partitioning Polytope Yong-Hong Kuo, Research Assistant Professor, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong - PRC, yhkuo@cuhk.edu.hk, Janny Leung We study the polyhedral structure of the mixed set covering, packing and partitioning problem, derive the mixed odd hole inequality, and identify sufficient condition for it to be facet-defining. In the special case when the induced graph is a (mixed) odd hole, the inclusion of this new facet-defining inequality provides a complete polyhedral characterization. Computational experiments show that these new valid inequalities achieve a significant time reduction in solving the mixed problems. 4 - Using Big-Data Analytics for Identifying Hot Spots of Border Security

Conic Convex Optimization: New Algorithms and Results Sponsor: Optimization/Linear and Conic Optimization Sponsored Session

Chair: Robert Freund, Professor, MIT Sloan School of Management, Building E62-567, 77 Massachusetts Ave., Cambridge, MA, 02139, United States of America, rfreund@mit.edu 1 - Approximation Schemes for Linear Programming in Inner Product Spaces Sergei Chubanov, University of Siegen, Kohlbettstr. 15, Siegen, Germany, sergei.chubanov@uni-siegen.de The size of an LP is the sum of binary sizes of the coefficients describing this LP. Such LPs are known to be polynomially solvable. The situation changes if each element of the data can be accessed only via an oracle and the size is defined as the size of the data used by the oracle. This class includes dynamic flows and DP formulations of some other NP-hard problems. A further generalization leads to LPs in inner product spaces. In this talk, we discuss a new algorithm for such problems. 2 - Solving General Convex Conic Problems with First-order Methods James Renegar, Professor, Cornell University, 224 Rhodes Hall, Ithaca, NY, 14853, United States of America, renegar@cornell.edu We present recent results in ongoing research pertaining to a framework that is novel in allowing any convex, conic optimization problem to be recast as an equivalent convex optimization problem whose only constraints are linear equations and whose objective function has Lipschitz constant no greater than one, to which a broad class of first-order methods can be applied. 3 - New Computational Guarantees for First-order Methods for Convex Optimization, via a Function Growth Constant Robert Freund, Professor, MIT Sloan School of Management, Building E62-567, 77 Massachusetts Ave., Cambridge, MA, 02139, United States of America, rfreund@mit.edu, Haihao Lu We present new algorithms and complexity bounds for solving convex optimization problems using first-order methods. We presume we are given a strict lower bound on f^*. We introduce a new functional measure called the growth constant G for f(x) that measures how quickly the function level sets grow and that plays a fundamental role in the complexity analysis. We present new computational guarantees for non-smooth and smooth optimization that improves on existing complexity bounds in many ways. MA17 17-Franklin 7, Marriott Network Optimization under Uncertainties Sponsor: Optimization/Network Optimization Sponsored Session Chair: Neng Fan, University of Arizona, 1127 E. James E. Rogers Way Room 111, Tucson, AZ, 85721, United States of America, nfan@email.arizona.edu 1 - Identifying Critical Nodes of Interdependent Networks by Integer Programming Shanshan Hou, University of Arizona, Tucson, AZ, 85721, United States of America, shanshanh@email.arizona.edu, Neng Fan, Andres Garrido In this talk, we analyze the vulnerability of interdependent networks by identify a set of nodes in power grid, whose removal results high impacts by the cascading failures in the interdependent communication network and itself. We propose an approach by integer programming to identify such set of nodes. Knowing the behavior of these networks can help to be more prepared before attacks and failures that may affect the power network supply and functionality. 2 - Improving the Global Pre-positioning Network for Natural Disaster Recovery Adam Prokop, Graduate Msc Student In Supply Chain Management, Wilfrid Laurier University, 75 University Avenue West, Waterloo, ON, N2L3C5, Canada, prok3910@mylaurier.ca Pre-positioning critical supplies in strategic locations can increase the effectiveness of humanitarian relief aid for natural disasters. An optimization model, utilizing recent global disaster risk indexes, was developed to evaluate the current United Nations Humanitarian Response Depot network. Alterations of the current network were shown to significantly minimize the average distance between pre- positioning facilities and demand regions.

Haibo Wang, Killam Distinguished Asso Prof, Texas A&M International University, 5201 University Blvd, Laredo, TX, United States of America, hwang@tamiu.edu, Yaquan Xu, Jun Huang, Wei Wang

This project develops a comprehensive data aggregation and analysis system to provide the decision support for identifying hot spots of border security using a complex network model for transportation infrastructure in the border region. All these research related data will be aggregated on both space and time dimensions and analyzed by using “big data” models and tools developed in this study

150