Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WC08

3 - Combining Exact Global Optimization and Nature-inspired Meta- heuristics: Revealing Robust Black-box Optimization Problem Solving Characteristics Logan Mathesen, Research Assistant, Arizona State University, 699 S. Mill Ave, Tempe, AZ, 85281, United States, Giulia Pedrielli Stochastic Optimization with Adaptive Restart (SOAR), recently proposed by the authors, mixes local and global search through controlled restarts. Despite good empirical performance SOAR is difficult to effectively parallelize. In this work we treat SOAR as an agent, parallelizing, we mix explorative (L vy) agents with exploitative SOAR agents. Meta-models blend information of each agent type, improving SOAR agent search. The distributed SOAR (dSOAR) algorithm results; experimentation investigates relationships between agent types, revealing robust cost effective dSOAR optimization behavior. n WC08 North Bldg 124A Stochastic Optimization Methods and Approximation Theory in Machine Learning II Sponsored: Optimization/Nonlinear Programming Sponsored Session Chair: Aritra Dutta, Thuwal, 23955-6900, Saudi Arabia Chair: El houcine Bergou, INRA-KAUST, Jeddah, Saudi Arabia 1 - Random Direct Search Method for Minimizing Nonconvex, Convex and Strongly Convex Functions El houcine Bergou, INRA-KAUST, King Abdullah University of Science and Techn, Jeddah, Saudi Arabia In this paper we consider the problem of unconstrained minimization of a smooth function in a setting where only function evaluations are possible. We design a novel randomized direct search (RDS) method and analyze its complexity. At each iteration, RDS generates a random search direction according to a certain fixed probability law. We analyze RDS method under several stepsize selection schemes (fixed, decreasing, estimated through nite differences, etc). While deterministic direct search depends quadratically on n (n is the dimension of the space), our method depends linearly on n. We also propose a parallel Aritra Dutta, King Abdullah University of Science and Technology, Division of Computer, Electrical and Mathemat, Al Khwarizmi Bldg 1, Thuwal, 23955-6900, Saudi Arabia, Filip Hanzely, Peter Richtarik Robust principal component analysis (RPCA) is a well-studied problem with the goal of decomposing a matrix into the sum of low-rank and sparse components. In this paper, we propose a nonconvex feasibility reformulation of RPCA problem and apply an alternating projection method to solve it. To the best of our knowledge, we are the first to propose a method that solves RPCA problem without considering any objective function, convex relaxation, or surrogate convex constraints. We demonstrate through extensive numerical experiments on a variety of applications, including shadow removal, background estimation, face detection, and galaxy evolution, that our approach matches and often significantly outperforms current state-of-the-art in various ways. 3 - An Inexact Regularized Stochastic Newton Method for Nonconvex Optimization Xi He, Lehigh University, 837 Cedar Hill Drive, Allentwon, PA, 18109, United States We consider the minimization of nonconvex functions that typically arise in machine learning. In this paper, an inexact regularized stochastic Newton method (irSNT) within trust region scheme is purposed and it only requires Hessian- vector product oracle within the subproblem solver. irSNT achieves the best worst-case complexity regarding first-order stationary point convergence in terms of expectation and also has a second-order stationary point convergence guarantee with high probability. The Numerical experiments show that our proposed method can successfully solve optimization problems involving nonconvex objectives. 4 - Improved Shrinkage Prediction under a Spiked Covariance Structure Trambak Banerjee, University of Southern California, Los Angeles, CA, United States We develop a novel shrinkage rule for prediction in a high dimensional non- exchangeable hierarchical Gaussian model with an unknown spiked covariance structure. We propose a family of commutative priors which, governed by a power hyper-parameter, ranges from perfect independence to highly dependent scenarios. It induces a wide class of predictors whose evaluation involves version for RDS, with better iteration complexity bounds. 2 - A Nonconvex Projection Method for Robust PCA

quadratic forms of smooth functions of the unknown covariance. We propose an efficient adaptive prediction procedure which outperforms factor model based plug-in predictors by using uniformly consistent estimators of the quadratic forms involved in the coordinate-wise shrinkage strategies. We further improve our predictor by introducing possible reduction in its variability through a novel coordinate-wise shrinkage policy that only uses covariance level information and can be adaptively tuned using the sample eigen structure of the high dimensional spiked covariance model. Simulation studies are conducted to show that in many settings the proposed method substantially improves the performance of traditional plug-in based shrinkage procedures which first estimate the covariance and thereafter optimize over the hyper-parameters. n WC09 North Bldg 124B Healthcare Supply Chain Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Liu Yang, Purdue University, New Albany, IN, 47150, United States 1 - Service Level Satisfaction during Shortages for Health Networks Erhun Kundakcioglu, Ozyegin University, Faculty of Engineering, Nisantepe mah. Orman sok., Istanbul, 34794, Turkey, Cem Bozkir In this study, we consider a system of healthcare providers, which face the same uncertain supply disruptions (e.g., regionwide, nationwide, or worldwide drug shortages). Each hospital observes a stochastic demand and if demanded drug is unavailable, patients leave and receive care in another hospital system. As these unavailabilities hurt the brand value of the hospital system, we propose an inventory sharing mechanism for hospitals to mitigate the effect of uncertain supply disruptions. We explore reactive versus proactive inventory sharing approaches by investigating the effect of inventory related parameters on the service level of the system. 2 - Improving Supply Chain Process Efficiency and Data Transparency Using Blockchain Technology Raja Jayaraman, Khalifa University, P.O. Box 127788, Abu Dhabi, United Arab Emirates, Mecit Can Emre Simsekler Supply chain logistics constitutes second largest expense for healthcare providers. The exponential growth, global sourcing of products, and increasing healthcare costs presents a compelling need to adopt innovative technology and real-time information sharing across the supply chain. The current system-wide inefficiencies in healthcare supply chains present an important and timely opportunity to improve the way transactions are recorded, stored and shared amongst stakeholders. In this talk I will present some of the significant challenges in supply chain data management, present several applications of blockchain technology along the product supply chain with potential pitfalls. 3 - Optimizing Medical Supply Spend for GPOs and Healthcare Organizations Liu Yang, Purdue University, Purdue Research Park, 3000 Technology Avenue, New Albany, IN, 47150, United States Tier pricing is commonly used by medical suppliers to support corporate sales and market share strategies. Hospitals and GPOs have the opportunity to achieve significant savings in medical supply costs if they can effectively utilize tier pricing, but a major challenge is that tier structures vary by vendors and product categories, and may be based on volume, spend, and/or market share, and could be at a single facility or across facilities. This research presents a modeling framework that addresses the complexity of tiers with the consideration of one- way cross-reference and preference of individual facility. The application enables hospitals to achieve over 15% reduction in supply costs. 4 - A Comparative Analysis of Healthcare and Traditional Supply Chains Using Financial Ratios Balaraman Rajan, California State University East Bay, 25800 Carlos Bee Blvd, Hayward, CA, 94542, United States, Vishwanath Hegde We compare and contrast the structure of healthcare and traditional supply chains using publicly available financial data. First, we develop a framework to analyze the healthcare supply chain and use the framework to compare with traditional supply chain. Then, by analyzing financial ratios, we find that companies that operate in different stages of the healthcare supply chain do exhibit different characteristics and differ from the companies that operate in comparable stages of traditional supply chains. We then draw inferences about the structural differences between healthcare supply chain and traditional supply chains based on the observed patterns.

472