2016 INFORMS Annual Meeting Program

SA86

INFORMS Nashville – 2016

Sunday, 11:00AM - 12:30PM

SA86 GIbson Board Room-Omni

Manufacturing I Contributed Session

SB01 101A-MCC

Chair: Mohsen Moghaddam, Postdoctoral Researcher, Purdue University, Grissom Hall, 315 N. Grant Street, West Lafayette, IN, 47907, United States, mmoghadd@purdue.edu 1 - Identifying Shifting Production Bottlenecks Using Clearing Functions Reha Uzsoy, Professor, North Carolina State University, Dept. of Industrial & Systems Engg, 300 Daniels Hall Camps Box 7906, Raleigh, NC, 27695-7906, United States, ruzsoy@ncsu.edu, Baris Kacar, Lars Moench Production planning models using clearing functions can provide meaningful dual prices for resources that are not fully utilized. We present a case study of the analysis of a semiconductor wafer fabrication system using this approach, and demonstrate the rapidly shifting nature of production bottlenecks even under stable demand. 2 - Determinants Of Commercial Exploitation For European Funded Technological R&D In Manufacturing Vasco Figueiredo Teles, Researcher, MIT Portugal Program, Porto, Portugal, vbteles@inesctec.pt, Abilio P. Pacheco, Abilio P. Pacheco, Joao Claro, Joao Claro A significant number of technologies resulting from R&D funded by the European Commission and aiming at commercial exploitation, do not achieve success in the marketplace, or in fact even reach it. We use regression analysis and a data set describing 60 technologies from European R&D projects in manufacturing, to identify potential determinants of exploitation. The technologies are classified in a 4-stage exploitation scale, and their characteristics (type, sector, geography, technology readiness level, or platform potential, among others) are compared among stages. Based on the identified determinants, we offer a set of suggestions on how to improve exploitation support in these contexts. Sunday Plenary Davidson Ballroom-MCC Cognitive Computing: From Breakthroughs in the Lab to Applications on the Field Plenary Session Chairs: Chanaka Edirisnghe, Rensselaer Polytechnic Institute, & Ed H. Kaplan, Yale University and INFORMS 2016 President 1 - Cognitive Computing: From Breakthroughs In The Lab To Applications on The Field Guru Banavar, Vice President, IBM Research, Watson Research Center, Yorktown Heights, NY, United States, banavar@us.ibm.com In the last decade, the availability of massive amounts of new data, the development of new machine learning technologies, and the availability of scalable computing infrastructure, have given rise to a new class of computing systems. These “Cognitive Systems” learn from data, reason from models, and interact naturally with us, to perform complex tasks better than either humans or machines can do by themselves. These tasks range from answering questions conversationally to extracting knowledge for discovering insights to evaluating options for difficult decisions. These cognitive systems are designed to create new partnerships between people and machines to augment and scale human expertise in every industry, from healthcare to financial services to education. This talk will provide an overview of cognitive computing, the technology breakthroughs that are enabling this trend, and the practical applications of this technology that are transforming every industry. Sunday, 10:00AM - 10:50AM

Machine Learning Sponsored: Data Mining Sponsored Session Chair: Cynthia Rudin, MIT, 100 Main Street, Cambridge, MA, 02142, United States, rudin@mit.edu 1 - Generalized Inverse Classification Michael Lash, University of Iowa, Iowa City, IA, United States, michael-lash@uiowa.edu, Qihang Lin, Nick Street, Jennifer Robinson, Jeffrey W Ohlmann Inverse classification (IC) is the process of perturbing a test point such that the predicted probability of a specific class is minimized. In previous work, we outlined an IC framework that incorporated a linear cost function and solved the problem by assuming the classifier was differentiable. In this talk we extend the framework to non-linear costs and relax our assumptions. We demonstrate that, using heuristic-based methods, the IC problem can be solved using arbitrary classifiers, about which only basic assumptions are made. 2 - On Difference Of Convex Optimization To Visualize a Word Storm In this talk we address the problem of visualizing in a bounded region a set of individuals, which has attached a dissimilarity measure and a statistical value. This problem, which extends the standard Multidimensional Scaling Analysis, is written as a global optimization problem whose objective is the difference of two convex functions (DC). Suitable DC decompositions allow us to use the DCA algorithm in a very efficient way. Our algorithmic approach is used to visualize a dynamic linguistic real-world dataset. 3 - Consensus Based Modeling Using Distributed Feature Construction Haimonti Dutta, University at Buffalo, haimonti@buffalo.edu Inductive Logic Programming can be used as a tool for discovering relational features for subsequent use in a predictive model. However, such models often do not scale. In this paper, we address this computational difficulty by allowing features and models to be constructed in a distributed manner. There is a network of computational units, each of which employs an ILP engine to construct a small number of features and build a (local) model. Then a consensus-based algorithm is learnt, in which neighboring nodes share information to update local models. For a category of models (those with convex loss functions), it can be shown that the algorithm will result in all nodes converging to a consensus model. 4 - Regulating Greed In Multi-Armed Contextual Bandits Stefano Traca, MIT, stet@mit.edu Abstract to come SB02 101B-MCC Data Mining in Medical and Sociological Decision Making Sponsored: Data Mining Sponsored Session Chair: Kamran Paynabar, Georgia Institute of Technology, Atlanta, GA, United States, kamip@umich.edu 1 - Single Stage Prediction With Text Data Using Dimension Reduction Techniques Shawn Mankad, Cornell, spm263@cornell.edu Text data is playing an increasingly important role within the business world for economic analyses and operations management. There are many ways to summarize and transform unstructured data into actionable insights. We compare several modern text analysis methods for prediction of economic outcomes to derive guidelines for researchers and practitioners. Dolores Romero Morales, Copenhagen Business School, drm.eco@cbs.dk, Emilio Carrizosa, Vanesa Guerrero

42