Plenaries & Keynotes

Daily Plenaries

Sunday, October 20


Plenary: Seeking Perfection: Reflections on the Journey

CC – 6E, Level 6

This session will review the Virginia Mason journey over 17 years to transform its organization and culture. From the early changes and challenges focused on old paradigms to the adoption of the Toyota Production System (Lean Management) as its management system, this session will highlight early resistance to change, the development of a new strategic plan, adoption of new compacts with physicians, leaders, and board members, and hard wiring improvements in quality and safety for sustainability.

The session will include the critical role of leadership and governance and the implications of change in a very dynamic marketplace.

Gary Kaplan

Virginia Mason Health System

Monday, October 21


Plenary: The Parametric Self-Dual Simplex Method — A Modern Perspective

CC – 6E, Level 6

Omega Rho Distinguished Lecture

The parametric self-dual simplex method (PSD) dates back to George Dantzig’s classic book published in 1963. Other variants of the simplex method have dominated the world of optimization over the years but the PSD variant has lots of attractive features both from an educational perspective and in the world of applications. In the talk I will define the method, discuss how it helps one understand more deeply duality theory and algorithm complexity and I will discuss some application areas including portfolio selection problems and LAD-Lasso problems that arise in modern machine learning problems.

Robert Vanderbei

Princeton University

Tuesday, October 22


Plenary: Planning Transportation Capacity at Amazon

CC – 6E, Level 6

Amazon has continued to speed up delivery while the number of shipped packages continues to grow, shortening the time it takes between when an item is ordered and the moment that item arrives on a customer’s doorstep. Prime customers experience faster delivery speeds, whether that comes from the broad selection for free two-day and one-day delivery from coast to coast, free same day delivery in select locations, or free two-hour delivery with Prime Now on tens of thousands of items. At the same time, Amazon’s fulfillment business – along with that of other online retailers – has increased the pressure on parcel delivery transportation capacity both in the US and overseas. Continuing to satisfy customers’ needs and increase delivery speed will require more efficient use of existing networks and the creation of additional capacity for package sortation, line haul trucking, air freight, and last mile delivery. The challenge of designing a network that can meet these dynamic needs requires us to develop technical solutions implemented through design tools that incorporate the latest process innovations. We must overcome both technical and organizational challenges to design and operate the logistics network, involving facility locations, inventory placement, facility operations, network connections, and scheduling. While the technical challenges are at least partially met through solving large-scale combinatorial optimization problems, the organizational challenges require innovative mechanisms that motivate network-wide cooperation while also allowing decentralized, scalable operation.
The presentation will cover the challenges that we face, the way we have formulated some of the problems, and some the challenges and opportunities that remain.

Russell Allgor


Sunday Keynotes


Data Science in Online Marketplaces: An OR Perspective

CC – 6B, Level 6

Online marketplaces have seen immense growth in the past decade, opening the door to novel problems in prediction, control and optimization under the umbrella of Data Science. In this talk, we will focus our attention on pricing and matching problems in online marketplaces, drawing on examples at Uber Freight — an online marketplace for freight transportation. We will highlight the connection of these problems to operations research and economics.

Bar Ifrach

Uber Freight

Analytics at the University of Cincinnati: A History of Innovation and Practical Impact

CC – 6C, Level 6

UPS George D. Smith Prize

The Operations, Business Analytics & Information Systems (OBAIS) Department in the Carl H. Lindner College of Business at the University of Cincinnati has a proud history of impact on the practice of analytics. For more than 50 years, the OBAIS Department has partnered with industry, produced thought-leading research, and influenced the teaching of analytics. In this presentation, we will provide an overview of the historical and current innovative activities undertaken by the OBAIS Department that culminated in its receiving the 2019 INFORMS UPS George D. Smith Prize.

Michael Fry

University of Cincinnati

Glenn Wegryn, CAP

University of Cincinnati

Structured Optimization for Dexterous Robots

CC – 6A, Level 6

Despite the incredible apparent success of robotics on youtube, robots are still surprisingly bad at some tasks that we expect them to do well — for example, robots are still relatively inept when it comes to manipulating objects with their hands. A major challenge in robotic manipulation is the diversity of the desired tasks: how do we rigorously formulate the problem of chopping vegetables, or putting away the dishes? Taking an optimization-based approach, then what is the cost function and what are the constraints? Can we even evaluate the performance of the system before we have developed a perception system that is capable of interpreting raw camera images to understand the number and location of various vegetables and/or plates in the scene?

In this talk, I’ll describe our attempts to bring rigorous tools from optimization to bear on the problem of “open-world” robot manipulation. I will describe the non-smooth mechanics of making and breaking contact between the robot and the world, and how planning and control for non-smooth mechanics can be formulated using combinatorial+convex optimization like mixed-integer programming and/or semidefinite relaxations. I’ll describe the deep learning approaches that we are using for perception and learning “intuitive physics”. And I’ll describe some of the challenges of verifying learning autonomous systems of this complexity, including calibrating the complex distributions (with categorical and continuous random variables) that we use to describe the distribution over environments.

Russell Tedrake

MIT EECS / Toyota Research Institute

Monday Keynotes


Alleviating Poverty & Inequity: Fresh Challenges for Business Model Innovation

CC – 6C, Level 6

Poverty and inequity remain two of the most pressing challenges in today’s world. Meanwhile, business model innovation is transforming business across the globe, creating new wealth by combining innovative operational thinking, data and analytics to unleash previously untapped sources of value.

This talk will examine the opportunity for business model innovation to alleviate poverty and inequity, showing how our community can play an important role. I will discuss how business model innovations can alleviate poverty by simultaneously empowering women, increasing incomes, and creating low-cost, high-quality and environmentally sustainable products or services. I will highlight opportunities for high-impact research that require different forms of modelling expertise, including econometrics and data science.

Kamalini Ramdas

London Business School

The New Age of Healthcare Delivery

CC – 6B, Level 6

MSOM Fellow

The U.S. healthcare system is undergoing one of the most rapid periods of change ever. The confluence of new payment systems, technological innovations, greater data availability and higher consumer expectations is resulting in a new focus on increasing access and convenience, particularly for high-risk, high-needs patients, while controlling costs through greater reliance on evidence-based care and increasing productivity. While hospitals and physicians have dominated the healthcare delivery system in the past, new models of healthcare delivery employ a variety of locations and types of providers, raising questions about the best means to deliver cost-effective care. In this talk I’ll describe these operational and organizational innovations and the critical research questions that must be addressed to create higher quality, lower cost healthcare systems in the future.

Linda Green

Columbia Business School

Zero Carbon Analytics

CC – 6A, Level 6

IFORS Distinguished Lecturer

Climate change from global warming is the most critical problem currently facing mankind. The deleterious effects of global warming have already been felt in 2019 in terms of storms, heatwaves and forest fires. Climate-based natural disasters in the longer term threaten to be much worse. Policies to deal with this problem must combine efforts to reduce global warming by eliminating greenhouse gas emissions with actions that adapt to worsening climate outcomes. Determining the best policy settings is a highly complex decision problem, involving complex interacting systems and markets, many stakeholders, many decision criteria, and huge uncertainty. In other words, it is a major opportunity for operations researchers.

This talk will illustrate some of the analytics challenges in planning a net-zero carbon economy with a focus on electricity systems. Models range from short-term operations (like dealing with intermittent renewable energy) to long-term investment planning to ensure security of energy supply in an uncertain future. Planning models are complemented with equilibrium models that represent the effects of risk-averse agent behaviour. The models will be illustrated with a New Zealand case study based on our recent work for the New Zealand Interim Climate Change Committee in planning a 100% renewable electricity system.

Andy Philpott

University of Auckland

INFORMS AI Strategy: Opportunities at the Intersection of AI and Operations Research

CC – 6E, Level 6

From driverless cars to deep learning, from AlphaGo to poker, from health care diagnosis and delivery to smart operations, applications of Artificial Intelligence (AI) have captured public imagination and become an active topic of discussion among leaders in all sectors. The Trump administration released an executive order on AI (see and held a White House summit in September 2019 on “Public Sector Implementation of AI“. As a community and discipline that has long focused on developing solutions to important societal and business decision problems, what should our posture as a society be to these developments in AI? In this keynote presentation, we will present the key ideas underlying the AI strategy for INFORMS and the work already underway that will find ways for members to benefit from the rise of AI. This will draw on the work done by the board, the AI strategy committee, and countless members of INFORMS.

Ramayya Krishnan

INFORMS President
Heinz College of Information Systems and Public Policy
Carnegie Mellon University

Pascal Van Hentenryk

Co-Chair, AI Strategy Committee, INFORMS
ISYE, Georgia Tech University

Tuesday Keynotes


Learning Personalized Policies: Theory and Applications

CC – 6E, Level 6

An important application of machine learning concerns the problem of learning treatment assignment policies that assign the optimal treatment to individuals as a function of their observable characteristics. This talk will review recent developments in offline policy learning and hypothesis testing, whereby historical data is used to estimate optimal policies accounting for the fact that personalization and adaptive experimentation or bandits were used in historical treatment assignments. A variety of empirical applications will be presented, including applications for personalized pricing, contextual bandit experiments, and targeted government policies.

Susan Athey

Stanford Business School

Predictive Data Science for Physical Systems: From Physics-based Models to Scientific Machine Learning

CC – 6C, Level 6

For high-consequence decisions in science, engineering and medicine–big decisions–we need more than just big data, we need big models too. These problems are characterized by complex multi-scale multi-physics dynamics, where small changes in parameters can lead to drastic changes in system behavior. They are also typically characterized by high-dimensional uncertain parameters that cannot be observed directly, and by a need to issue predictions that go well beyond the specific conditions where data may be available. For these reasons, a purely data-focused perspective will fall short. Achieving predictive data science for these complex physical systems requires a synergistic combination of data and physics-based models. Learning from data through the lens of models is a way to bring structure to an otherwise intractable problem: it is a way to respect physical constraints, to embed domain knowledge, to bring interpretability to results, and to endow the resulting predictions with quantified uncertainties. As one example, this talk shows how formulations from projection-based model reduction can be combined with machine learning methods to achieve this. Case studies in aerospace engineering applications demonstrate the importance of embedding physical constraints within learned models, and also highlight the important point that the amount of model training data available in an engineering setting is often much less than it is in other machine learning applications, making it essential to incorporate knowledge from physical models.

Karen Willcox

University of Texas Austin

Edelman Reprise – Protecting Community Waterways: Applying Analytics and Optimization for Wastewater Management

CC – 6B, Level 6

The Louisville and Jefferson County Metropolitan Sewer District (Louisville MSD) in Kentucky, USA, uses operations research, advanced analytics and innovative engineering concepts to protect local community waterways by optimizing the collection, transport and treatment of wastewater. In partnership with Tetra Tech, Louisville MSD pioneered the application of real time control using Csoft®, which relies on sewer monitoring data, weather forecasting, and data analytics for system-wide optimization. In operation since 2006, the solution has reduced more than 2 billion gallons of sewer overflows per typical year and saved the community over $200 million of capital cost.

Angela Akridge

Louisville Metropolitan Sewer District

Wolffie Miller

Louisville Metropolitan Sewer District

Diana Qing Tao

Tetra Tech

Wagner Prize Winner

CC – 6A, Level 6

The Daniel H. Wagner Prize is awarded for a paper and presentation that describe a real-world, successful application of operations research or advanced analytics. The prize criteria emphasize innovative, elegant mathematical modeling and clear exposition.

Wednesday Keynotes


People, Machines, and Intelligence

CC – 6C, Level 6

Advances in capabilities and applications of machine learning have brought to the fore multiple challenges and opportunities with human-AI collaboration. I will present research on principles and mechanisms for harnessing the complementary skills of people and machines. Studies in this area show the promise of aiming machine learning, decision making, and optimization at the challenge of weaving together human and machine intellect. I will describe efforts on guiding the display of information, controlling a mix of human and machine initiatives, fusing human and machine contributions, and shaping the construction of predictive models based on human capabilities. I will introduce key ideas and solutions in the context of efforts in transportation, aerospace, healthcare, and productivity applications.

Eric Horvitz


Marketplace Modeling: Managing Scale and Accuracy

CC – 6B, Level 6

Marketplace modeling is a research area that has had widespread real-world impact within the OR/MS community. Focusing on scale and accuracy, we will discuss some techniques we have been using to enable better decision-making for concrete applications, such as matching jobs and applicants, blood donation recommendations, friend suggestions, rendering feeds, and online advertising. These techniques include using competitive equilibria to compute allocations, finding compact representations of marketplaces, and using Bayesian optimization coupled with simulation to improve model accuracy.

Nicolas Stier-Moses