“It is my objective to generate knowledge that affects what the next generations learn,” begins Dimitris Bertsimas, former co-director of the MIT ORC and the Institute’s Boeing Leaders for Global Operations Professor of Management. “Is it impactful? That’s how I judge what is important.”
It is this philosophy of teaching and researching that has driven Bertsimas for decades. His current research focuses on three areas:
The first, personalized medicine, grew out of personal tragedy, when Bertsimas lost both of his parents to disease in 2009. After experiencing firsthand the limitations and variations of medical treatments available, he set out to apply an analytics approach to treating human diseases, such as diabetes and cancer, based on extensive medical data. In 2013, this research by his team—comprised mainly of ORC PhD students—won the best paper award in health care from the INFORMS professional society.
Bertsimas expects that one day doctors will prescribe specialized treatments based on an algorithm that looks at patients’ historical data and compares it to that of millions of others. “I believe this [research] will change how disease is treated today,” he states.
Bertsimas’ second area of research is robust optimization, the goal of which is to propose a tractable theory for optimization under uncertainty—take air traffic control, for example where the uncertainty is mainly to the weather. Over the past 15 years, he has worked with a range of ORC students—many of whom are now professors—developing optimization methods that can be applied to such diverse areas as energy, finance, and health care. The key contribution is a tractable theory of optimization under uncertainty. He has been teaching a doctoral class entitled "Robust Optimization" for the last several years based on his work in the area.
His final research area involves bringing the fields of operations research—particularly optimization and statistics—together. The key plan is to revisit some of the central problems of statistics and machine learning under a modern optimization lens that includes discrete, robust and convex optimization. Examples include sparse regression, optimal and robust classification, optimal factor analysis, sparse principal component analysis, sparse estimation of inverse covariance matrices among many others. He has introduced a doctoral class entitled "Statistical Learning under a Modern Optimization Lens" in the Spring 2016.
At the ORC, Bertsimas is fulfilling his goal to make a positive impact on the field—and on the world. He has worked with dozens of ORC students as teacher, mentor—and, in a few cases, best man at their weddings. He has created courses based on his research in statistics, analytics, and robust optimization. And he has authored or co-authored hundreds of scholarly works, including his recent book, The Analytics Edge, which he aptly dedicates to his “past, present and future doctoral students.”
Vivek Farias did not seek out a career in operations research (OR); instead, the electrical engineer “fell into” the field while in graduate school, when he needed to use OR tools to conduct his research.
“I found more and more that OR just happened to be a name for a bunch of things that I thought were very valuable,” he recalls. “Here’s a set of tools, ideas, and problems that are broadly applicable to the world. I found that notion really exciting.”
Now the Robert N. Noyce Career Development Associate Professor at MIT, Farias teaches and researches OR topics, specifically dynamic optimization in the face of uncertainty—a concept that has enthralled him since his days as an undergraduate building record-beating computer players for the game of Tetris. <- Edited.
Farias’ research in approximate dynamic programming has tangible applications in such diverse fields as finance and health care. And his work in high dimensional statistics has far-reaching applications in marketing, e-commerce, and retail—in fact, it led to his co-founding a company, Celect, which offers retail analytics for merchandise planning and assortment optimization; he currently serves as Celect’s chief technology officer.
Farias is quick to point out how his entrepreneurship enhances his teaching and researching at the ORC.
“I’m exposed to the actual problems — as opposed to problems I make up in my office. This informs the research I’m doing with students, and they benefit from these real-world scenarios.”
For Retsef Levi, the J. Spencer Standish (1945) Professor of Operations Management at MIT, one of the most significant features of the ORC is its intense focus on research—from the very start.
A member of the ORC since 2006, Levi stresses the importance of the unique, hands-on work students are able to do in the Center, often in collaboration with industry partners, like hospitals and government agencies.
“[Industry collaboration] allows me and my students to not only access real data and real problems, but also develop models and inform decisions that matter in the real world,” he explains.
For example, in 2011, Massachusetts General Hospital implemented a dramatic change to its surgical schedules based on modeling and analytics conducted by Levi and his students. The hospital also applied a scheduling algorithm they developed to smooth the utilization of infusion chairs throughout the day; both changes will ultimately lead to better patient care.
Currently, Levi is working with ORC students and faculty to assess and mitigate risks in the global food supply chain. According to him, 40 million food shipments arrive at U.S. borders every year—and the government only has the resources to check a very small number of those.
Through data-driven analytics and state-of-the-art modeling, his team aims to predict which companies are likely to be involved in economically motivated (intentional) adulteration of food for financial advantage.
“At the ORC, our students have the opportunity to potentially inform how companies and regulatory bodies will act upon risk,” says Levi. “That’s an opportunity that very few places could offer to their students.”
“The ORC brings together leading researchers and exceptionally talented students under the same umbrella,” describes Rahul Mazumder, assistant professor of operations research and statistics at MIT. “They [are] using innovative scientific approaches to tackle some of the most complex problems in the real world.”
To that end, Mazumder, a self-described applied statistician doing research at the intersection of statistical machine learning and computational optimization motivated by real-world problems; is currently working on research that has applications in climate science and global sustainability.
“The main goal in this project is to harness massive amounts of environmental data gathered by automated floats deployed in the worlds’ oceans,” explains Mazumder, who is collaborating with researchers at Columbia University. “This data helps us understand how environmental parameters like ocean water salinity and nutrient content change over space and time—and this subsequently helps us understand important issues like climatic variations and environmental impact on the worlds’ fisheries.”
Mazumder, who joined the ORC faculty in 2015, finds himself in awe of the Center’s unique blend of collaboration and creativity—both in and out of the classroom.
“I get to be a part of a vibrant community of faculty and students, who are constantly striving to change the world for the better with a combination of research and education,” he enthuses. “It is an honor and a privilege to be a part of this energetic, intellectually curious group of people.”
“We believe in having an impact,” explains Georgia Perakis, the William F. Pounds Professor of Management and Codirector of the Operations Research Center at MIT. “That’s a distinguishing feature of the research we do here at the ORC … and it’s what motivates me as a professor.”
Indeed, since joining the MIT faculty in 1998, Perakis has spearheaded many significant projects in operations research—in areas ranging from energy to retail among others.
On the retail side, Perakis’ research focuses on helping companies run promotions and markdowns more efficiently. She believes in using data to build and analyze models that solve real problems and have an impact in the retail space. Using optimization and machine learning, she and her PhD students from the Operations Research Center together with collaborators from the retail industry, are building key models that can impact not only the development of predictive and prescriptive analytics, but also improving retail practice. For example, together with industry collaborators, they are building and analyzing models for promotion planning and pricing as well as models for personalized pricing and product recommendations acknowledging that a “one-size-fits-all-approach” does not work in today’s world.
In the energy sector, Perakis works closely with master’s and ORC PhD students to predict where damage could occur by superstorms or by corrosion and gas leaks. By analyzing data and building machine learning and optimization models, her research team builds models that enable utilities to predict damage from—and in response to—destructive events, helping them determine how many power restoration crews will be needed and where they should be placed before a storm hits, for example, and how to manage personnel so as to respond faster to gas leaks and avoid explosions.
Perakis is passionate about her research and its ability to meaningfully impact industry practice. But perhaps the greatest impact she has is on her students.
As evidenced by the decorative photo collage that adorns her office—given by and containing pictures of her current and former ORC students—the bonds she forms with her students are strong and long-lasting.
“That’s why I love my job,” says Perakis, gesturing to the collage. “They are extremely special to me. They’re my family.”