Informs Annual Meeting 2017

SD71

INFORMS Houston – 2017

5 - Time Series Classification using Normalized Subsequences Iman Vasheghani Farahani, North Carolina State University, 400 Daniels Hall, Raleigh, NC, 27695-7906, United States, ivasheg@ncsu.edu Iman Vasheghani Farahani, SAS.Institute Inc, 100 SAS.Campus Dr, Cary, NC, 27513, United States, ivasheg@ncsu.edu, Alex Chien, Russell Edward King In time series classification problems, finding normalized subsequences, whose distances from a time series indicates its label, requires a prior information on the length of the subsequence and a similarity measure between series. This may require a costly search be performed on the entire dataset. To address this problem, we examine the efficiency of data transformation using symbolic aggregate approximation. This transformation keeps the local structure of a time series, that enables us to find similar subsequences repeated in samples time series. We use decision trees classifiers and information gain as the criterion to grow the tree. 371F Journal of Global Optimization: INFORMS Issue Sponsored: Optimization, Global Optimization Sponsored Session Chair: Sergiy Butenko, Texas A&M University, College Station, TX, 77843-3131, United States, butenko@tamu.edu 1 - Parallel Distributed Block Coordinate Descent Methods Based on Pairwise Comparison Oracle Kota Matsui, Nagoya University, Nagoya, Japan, matsui.k@med.nagoya-u.ac.jp, Wataru Kumagai, Takafumi Kanamori We present a block coordinate descent method to solve unconstrained optimization problems. Our algorithm uses only pairwise comparison of function values, which tells only the order of function values over two points. In our algorithm,a Newton-type search direction is estimated and then a numerical solution is updated along the direction. The computation in the direction estimate can be easily parallelized, and thus, the algorithm works efficiently. Also, we theoretically derive an upper bound of the convergence rate for our algorithm and show that it achieves the optimal query complexity for specific cases. In numerical experiments, we show that our method outperforms some existing methods. 2 - Geodesic and Contour Optimization using Conformal Mapping SD71 We devise the shadow casting scheme, in which global basins of attraction are mapped to a much larger space and recovered by a reverse mapping. Under certain criteria, the enlarged space can be as larger the search space. We also introduce a novel trajectory method that couples to the shadow casting scheme. Local search methods such as Newton-Ralphson are employed to find the global optimum. Experimental results indicate that our proposed method can outperform recent derivative-free DIRECT variants in number of function/gradient evaluations on functions with many optima or when the global optimum is very close to a local one. 3 - Igreen: Green Scheduling for Peak Demand Minimization Shaojie Tang, TX, United States, Shaojie.Tang@utdallas.edu, Jing Yuan, Zhao Zhang, Ding-zhu Du Home owners are typically charged differently when they consume power at different periods within a day. Specifically, they are charged more during peak periods. Thus, in this paper, we explore how scheduling algorithms can be designed to minimize the peak energy consumption of a group of homes served by the same substation. We study both offline and online models. 4 - Solving Linear Optimization Over Arithmetic Constraint Formula Li Chen, Institute of Software, Chinese Academy of Sciences, Beijing, China, chenli@nfs.iscas.ac.cn, Yongji Wang Since Balas extended linear programming problem to disjunctive programming (DP) problem, researchers explored this optimization under various scenarios such as generalized disjunctive programming (GDP) and optimization modulo theories (OMT). In this paper, we convert linear DP/GDP model and linear- arithmetic OMT problem into an equivalent form referred to as linear optimization over arithmetic constraint formula (LOACF). A branch-and-bound based algorithm named RS-LPT for LOACF is proposed. We evaluate RS-LPT against two DP/GDP methods, three OMT solvers, and disjunctive transformation based method on different benchmarks. Our results favor RS-LPT especially for large scale cases. Ricky Fok, York University, Toronto, ON, Canada, ricky.fok3@gmail.com, Aijun An, Xiaogang Wang

5 - Incorporation of Delivery Times in Stereotactic Radiosurgery Treatment Optimization Dionne Aleman, University of Toronto, 5 King’s College Road, Toronto, ON, M5S.3G8, Canada, aleman@mie.utoronto.ca, Hamid R.Ghaffari, David A.Jaffray, Mark Ruschin Although mathematical optimization can greatly improve the quality of treatment plans in radiation therapy, one complication is the clinically unrealistic nature of optimized treatments. The difficulty arises from long treatment times and machine limitations that govern the minimum amount of radiation delivery time, which requires binary variables. We examine these issues within the penalty- based sector duration optimization (SDO) model for Elekta’s Leksell Gamma Knife Perfexion and the combined SDO and isocentre location model to reduce beam- on time and to ensure machine limitations are met. 372A Network Flow Problems in Applications Sponsored: Optimization, Network Optimization Sponsored Session Chair: Alexander Nikolaev, University at Buffalo, Buffalo, NY, 14260- 2050, United States, anikolae@buffalo.edu 1 - Monotonicity and Conformality in Multicommodity Network-flow Problems Frieda Granot, Sauder School of Business, Vancouver, BC, Canada, frieda.granot@sauder.ubc.ca We present a monotonicity theory for the important class of minimum convex- cost parametric multicommodity network-flow problems defined over general graphs. The results determine when it is possible to predict, without numerical computations, the direction of change of optimal multicommodity flows resulting from changes in arc-commodity parameters. In particular, we provide necessary and sufficient conditions that for every cost function satisfying a submodularity- convexity-compactness hypothesis there exists an optimal multicommodity flow for which the flow of a commodity in a given arc a is nondecreasing (resp., nonincreasing) in the parameter of a distinct commodity in arc b. 2 - A Mesoscopic Model to Evaluate the Evacuation Dynamics in a University Campus Jorge Huertas, Universidad de los Andes, Calle 115 No 55 C.- 54, Bogota, 11111, Colombia, ja.huertas1845@uniandes.edu.co, Jony A. Zambrrano, Andres Medaglia In this work we integrate a macroscopic and a microscopic model to evaluate the evacuation dynamics in a university campus. The macroscopic approach is a network optimization model that determines the optimal evacuation paths. The microscopic model is a discrete event simulation that models the congestion on the paths determined by the first model. We propose an iterative process in which the output of one model is the input of the other one, achieving a convergence on the evacuation time. The result is a mesoscopic model that considers bottlenecks, direction of flows, speed fluctuations, and congestion during an evacuation. Finally, we compare our results with real evacuation exercises in the campus. 3 - Reach Maximization for Direct Incentive-Driven Pro- Environmental Social Programs Alexander Nikolaev, University at Buffalo, 409 Bell Hall, Buffalo, NY, 14260-2050, United States, anikolae@buffalo.edu, Jose Walteros While exploiting the power of network-based social influence for the societal benefit is a breakthrough in itself, further research is required to boost the impact of such programs with direct incentives. We present investigations into the design and planning of randomized incentive programs — social lotteries — that inform consumers of their peers behaving pro-environmentally while being rewarded for doing so. We conduct calculated maximization of the ‘’reach’’ of such a program or policy, exploiting column generation in mixed-integer programming. SD72

126