Informs Annual Meeting 2017

MA74

INFORMS Houston – 2017

2 - Probabilistic Power Dominating Set Problem Ou Sun, University of Arizona, 827 E Drachman St, Tucson, AZ, 85719, United States, suno@email.arizona.edu, Neng Fan Power dominating set problem is widely used in power system. An important application in this area is the minimum Phasor Measurement Unit (PMU) Placement Problem which has zero-injection bus property. However, after sensor placement, there may be some unexpected failures occur on PMUs or transmission lines, which can affect the function of whole power system. Starting from this idea, we propose two methods to formulate the probabilistic coverage model under uncertainty of PMUs and lines given covering probability of each bus. One is changing the chance constraint into logarithm, another is calculating the covering probability of each bus in the network design graph. 3 - Optimization of Microgrid Operations with Renewable Energy Integration Jose L.Ruiz Duarte, University of Arizona, Tucson, AZ, 85721, United States, jlruizduarte@email.arizona.edu, Neng Fan The development and integration of new technologies to the traditional power grid are transforming it into a smart grid capable of responding to electricity demand growth and integration of renewable energy sources. Microgrids are important part of this approach. In this presentation, a mixed integer linear program is developed for operations of a microgrid, including unit commitment decisions, line connectivity decisions with network topology, power trade with main grid, islanding operations, energy storage systems, and solar energy integration. The uncertainty of solar power is captured by robust optimization, and problem is solved by Column-and-Constraint Generation algorithm. 372B Ethics in AI and OR Invited: Operations Research and the Future of Computing Invited Session Chair: Tae Wan Kim, PHD, Carnegie Mellon University, Posner #243, 500 Forbes Avenue, Pittsburgh, PA, 15213, United States, twkim@andrew.cmu.edu 1 - Autonomous Machines are Ethical John Hooker, Carnegie Mellon University, Tepper School of Business, Pittsburgh, PA, 15213, United States, jh38@andrew.cmu.edu While many see the prospect of autonomous machines as threatening, autonomy may be exactly what we want in a superintelligent machine. There is a sense of autonomy, deeply rooted in the ethical literature, in which an autonomous machine is necessarily an ethical one. Development of the theory underlying this idea not only reveals the advantages of autonomy, but it sheds light on a number of issues in the ethics of AI. It helps us to understand what sort of obligations we owe to machines, what obligations they owe to us, and whether to assign responsibility to machines or their creators. 2 - The Missing Ingredient in Robotics and Artificial Intelligence: Intrinsic Values Thomas Donaldson, PhD, The Wharton Schoo l, University of Pennsylvania, Philadelphia, PA, United States, donaldst@wharton.upenn.edu The rise of robotics and AI will force management organizations to take intrinsic values more seriously. Intrinsic values are non-derivative, impartial and synoptic. Non-intrinsic values, in contrast, are derivative, meaning that they derive either from mere postulation or from the perceptions of individual agents. New AI/robotics challenges frustrate traditional analyses that tend to use non-intrinsic values. 3 - Ethics and Accountability of Algorithms Kirsten Martin, PhD, George Washington University, martink@email.gwu.edu Recent headlines warn against the hidden and unchecked biases of algorithms used in advertising, sentencing, hiring, lending, etc. Hiding behind the apron of these headlines is a series of critical decisions about the design of the algorithm and the role of algorithms in decision making. The goal of this presentation is to explore how algorithms, which are developed, used, and sold by firms, can be biased and unjust, perform an important role in ethical decisions, and have implications for accountability in corporate responsibility. This paper offers a framework to examine the value-laden decisions when utilizing an algorithm to augment organizational and individual decision-making. MA73

4 - Algorithmic Transparency, a Right to Explanation and Placing Trust

Tae Wan Kim, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, twkim@andrew.cmu.edu, Bryan Routledge

A growing number of researchers and governmental bodies have called for algorithmic accountability. European Parliament and Council adopted General Data Protection Regulation, which will come into force in 2018. Authors debate whether GDPR, in addition to a right to be forgotten, grants EU residents another novel kind of protection, a so-called “right to explanation.” Here we articulate a tripartite structure of the right, justify it, and explore possible models of an ex post explanation that companies can provide to data subjects. Since we explore the moral or ethical, not legal, foundations of a right to explanation, the implications of our work are beyond the European context. 372C Graph Partitioning Sponsored: Computing Sponsored Session Chair: Onur Seref, PhD, Virginia Tech, Virginia Tech, Blacksburg, VA, United States, seref@vt.edu 1 - Distributed Spectral Clustering and Applications Shahzad Bhatti, University of Illinois at Urbana Champaign, 104 S Mathews Ave, Urbana, IL, 61801, United States, bhatti2@illinois.edu, Carolyn L.Beck Clustering is fundamental to machine learning. Most clustering algorithms are tailored for data stored on a central machine and are inappropriate for large data distributed across machines. The scale of the data may also prohibit moving it to a central machine. Our approach employs two phases: 1) individual machines generate a set of representative points for local data and communicate these to a central machine; 2) the central machine performs spectral clustering on the representatives and communicates cluster assignments to the distributed machines. We have explored various algorithms to generate representatives and compared results. Our algorithm is easily cast in the MapReduce framework. 2 - An Extended Formulation of the Convex Recoloring Problem on a Tree Minseok Ryu, University of Michigan, 1535 Pine Valley Boulevard, Apt 212, Ann Arbor, MI, 48104, United States, msryu@umich.edu, Sunil Chopra, Bartosz Filipecki, Kangbok Lee, Sangho Shim, Mathieu Van Vyve We introduce a strong extended formulation for the convex recoloring problem on a tree, which has an application in analyzing phylogenetic trees. The extended formulation has only a polynomial number of constraints, but dominates the conventional formulation and the exponentially many valid inequalities introduced by Campelo et al. The solution time using the extended formulation is much smaller than that with the conventional formulation. Moreover, the extended formulation solves all the problem instances attempted in Campelo et al. (2016) and larger-sized instances at the root node of the branch-and-bound tree without branching. MA74 We explain how the open source CONJECTURING program can be used to generate conjectures for the integrality gap for IP problems. The conjectures are in the form of upper or lower bounds; these bounds are functions of user-input invariants that may come from the problem being modeled. We use the k-partite Balanced Tree Problem as an example of what is possible. The problem is to partition the vertices of a tree into k sets so that the number of edges with endpoints in different partitions is minimized. We present a variety of conjectured bounds, in terms of the order n, the size m, and other invariants, and discuss an iterative process of investigation—-which is applicable to any similar optimization problem. 5 - Balanced Tree Partitioning Paul Brooks, Virginia Commonwealth Univ, Dept of Stat Sci and OR, P.O. Box 843083, Richmond, VA, 23284, United States, jpbrooks@vcu.edu In this talk, we present two integer programming formulations for the k-Balanced Partitioning problem for trees. One is an adaptation of a formulation for all tree partitions. The other is a novel flow-based formulation with a smaller integrality gap. We describe solution strategies and present computational results comparing the formulations. 3 - Automated Conjecturing for Integrality Gap Bounds Craig Larson, Virginia Commonwealth University, VA, United States, clarson@vcu.edu

159