2015 Informs Annual Meeting

WD27

INFORMS Philadelphia – 2015

2 - Mixed-integer Programming Formulations for Partial-order Plans Buser Say, Student, University of Toronto, 5 King’s College Rd., ON M5S 3G8, Toronto, ON, M5S 3G8, Canada, buser.say@mail.utoronto.ca, Andre Augusto Cire, Chris Beck A partial-order plan (POP) is a set of actions associated with precedence constraints for which a goal is achieved in any total ordering of actions that respects the precedence constraints. POPs are more flexible than sequential plans since agents can dynamically commit to the ordering of certain actions at execution time. We investigate novel mixed-integer linear formulations to produce valid POPs from sequential plans, and compare their performance to state-of-the-art MaxSAT models. 3 - Fast Optimal Chance-constrained Scheduling under Uncertainty Brian Williams, Professor of Aeronautics and Astronautics, MIT CSAIL, 32 Vassar St, Cambridge, MA, 02139, United States of America, williams@csail.mit.edu, Cheng Fang, Andrew Wang Temporal uncertainty in large-scale logistics requires balancing lost efficiency through slack and costly replanning when deadlines are missed. This motivates a computational framework to quantify and bound the risk of violating schedule requirements. In this work, we decompose the problem into two subproblems: 1) optimal risk allocation; and 2) enforcing schedule requirements. This allows us to minimize conservatism while leveraging specialized solvers for each subproblem for fast solutions. 4 - Linear Optimization for Operator Counting in Automated Planning Automated planning is the search for paths in factored state spaces. The recently suggested operator-counting framework uses LP/IP optimization to unify, combine and explain connections between state-of-the-art planning techniques. We present an easily accessible introduction to planning and operator counting. LP/IP methods are a hot topic in planning with significant potential for research collaborations between OR and AI researchers. Florian Pommerening, University of Basel, Spiegelgasse 5, Basel, 4051, Switzerland, florian.pommerening@unibas.ch Chair: Dongdong Ge, Shanghai University of Finance and Economics, 100 Wudong Road, Shanghai, 200433, China, gedong78@163.com 1 - An Extended Cutting Plane Approach with PWL Approximation for Generalized Geometric Programming Yiduo Zhan, PhD Student, University of Central Florida, 12800 Pegasus Drive, P.O. Box 162993, Orlando, FL, 32816, United States of America, yzhan@knights.ucf.edu, Chung-Li Tseng, Qipeng Zheng We employ an extented cutting plane (ECP) approach combining with piecewise- linear (PWL) approximation to provide the global solution of GGP. In this approach, the constraints are separated by positive and negative terms. The negative terms are converted to mixed-integer linear constraints through PWL approximation. The partially linearized GGP becomes a mixed-integer nonlinear problem (MINLP). This MINLP is solved using ECP method. Numerical problems are tested and results are discussed. 2 - Solving Non-separable Quadratic Binary Programs using Projection and Surrogation Jaehwan Jeong, Assistant Professor, Radford University, Department of Management, P.O. Box 6954, Radford, VA, 24142, United States of America, jjeong5@radford.edu, Chanaka Edirisinghe A new approach is developed to solve non-separable quadratic programs with binary variables. First, separability is induced using a projection technique and non-convex relaxation of the binary variables. Then, using constraint surrogation, an iterative sequence of nonconvex separable quadratic knapsack programs are solved efficiently using our previous algorithms. Preliminary computations are provided. 3 - Stochastic PDE-constrained Optimization of Vibrations of a Plate under a Piecewise-linear Current Dmitry Chernikov, University of Iowa, 1010 W Benton St. #208F, In this work a two-stage stochastic PDE-constrained optimization framework is applied to the problem of vibration control of a thin composite plate in the presence of electromagnetic field. The electric current is assumed to be of a piecewise-linear form. We compute the gradient of the objective function using adjoint numerical differentiation method. The value of the objective function is calculated by solving the governing PDEs, and a black-box approach is used for the minimization problem. Iowa City, IA, 52246, United States of America, dmitry- chernikov@uiowa.edu, Pavlo Krokhmal, Olesya Zhupanska WD27 27-Room 404, Marriott Optimization Nonlinear Programming I Contributed Session

4 - Optimizing Blending Operations at the World’s Largest Coal Export Port Fabian Rigterink, The University of Newcastle, Australia, University Drive, Callaghan NSW, 2308, Australia, fabian.rigterink@newcastle.edu.au, Natashia Boland, Thomas Kalinowski The port of Newcastle, Australia, is the world’s largest coal export port. We model the supply chain’s medium- and long-term planning of blending operations as a time-expanded pooling problem. Using new multi-commodity flow formulations, we study the trade-off between continuous and discretized variables (NLP and MINLP). We evaluate the performance of unary, log unary and binary variable discretizations in an extensive computational study that concludes the talk. 5 - On the Complexity and Algorithms of Regularized Least Square Problems Dongdong Ge, Shanghai University of Finance and Economics, 100 Wudong Road, Shanghai, 200433, China, gedong78@163.com, Yinyu Ye, Zizhuo Wang, Hao Yin We show that finding a global optimal solution for the regularized least square problem is strong NP-Hard as long as the nonlinear penalty function is concave and non-decreasing. This result clarifies the complexity for a large class of regularized optimization problems studied in the recent statistics literature. Chair: Tianqin Shi, San Jose State University, One Washington Square, Business Tower 450, San Jose, CA, 95192, United States of America, tianqin.shi@sjsu.edu 1 - Group Decision Making: A Flexible Methodology Pascale Zaraté, Professor, Toulouse 1 Capitole University - IRIT, 2 rue du Doyen Gabriel Marty, Toulouse Cedex 9, 31042, France, pascale.zarate@irit.fr The specific benefice of a collective decision process mainly rests upon the possibility for the participants to confront their respective points of views. To this end, they must have cognitive and technical tools that ease the sharing their own preferences, while allowing keeping some information and feelings for their own. The paper presents the basis of such a flexible, cooperative decision making methodology. This methodology has been implemented in a GDSS called GRoUp Support. 2 - Identifying Patients at Risk using Fuzzy Logic John Zaleski, Chief Informatics Officer, Nuvon, Inc., 4801 S. The use of “big data” for decision making has been a growing area of investigation and usage in healthcare enterprises. This paper shows how fuzzy rules can be used to operate on data obtained from the point of care to assist in clinical decision making, with application to real-time data collection in medical surgical units. 3 - The Effects of Patent Extension and Pharmaceutical Stewardship Program on Green Pharmacy Tianqin Shi, San Jose State University, One Washington Square, Business Tower 450, San Jose, CA, 95192, United States of America, tianqin.shi@sjsu.edu, Dilip Chhajed, Nicholas Petruzzi The eco-toxicity arising from unused pharmaceuticals has drawn considerable attention. In this paper, an innovative pharmaceutical company faces price- dependent demand and decides whether to adopt green pharmacy in response to the regulatory policy as well as the competition from a generic company. A pharmaceutical company incurs a fixed cost to choose green pharmacy. We examine the impacts of two regulatory policies, patent extension and take-back regulation, on the choice of green pharmacy. Broad Street, Suite 120, Philadelphia, PA, 19112, United States of America, jzaleski@nuvon.com WD28 28-Room 405, Marriott Decision Analysis V Contributed Session

462