Optimization comes in many flavors
INFORMS Annual Meeting is a place we often find great examples of what other people achieve by the power of analytics. Every year I try to find at least a talk which gives me joy of seeing what is possible to do with optimization. This year I didn’t have to wait for too long to find one. In this post, I will share my opinions about the content I listened at SAS workshop on Saturday.
Before I begin, let me tell you that vendor workshops on Saturdays are great to learn about where the state-of-the-art is. Since many people still are traveling on this day, I feel like it is often overlooked by attendants. This year, I flew for INFORMS on Friday and had plenty of time to attend one. And even though I’m an intern at SAS for over a year, I learned a great deal about what others are doing besides my own work in their workshop.
SAS’ workshop start with an overview of what they offer for optimization. Then, it is followed by a tutorial where several examples are discussed. Besides providing the ability to model optimization problems, their toolbox enables users to model algorithms by using optimization tools. An example of decomposition algorithm is shown during the workshop. I think capabilities of a software is best to be learned from their creators.
Later, they explained how they expand their toolbox to provide services for their customers under the everchanging nature of analytics. Machine learning using SAS Optimization is discussed in great detail. Optimization is at the heart of several machine learning algorithms without a doubt. Throughout the last year I worked on Auto-tuning of machine learning algorithm parameters, also known as hyperparameter optimization.
The consulting group in the OR department shared a pro-bono project they worked. Pelin, my lovely wife, discussed how they worked with DataKind to solve the bus stop consolidation, fleet scheduling and bell time assignment problems of Boston Public Schools, which is the oldest school system in US. Their work resulted with a projected savings between 3 to 10 million dollars for BPS. The way how they approach the problem shows the group’s experience of dealing similar problems in the past. The workshop ended with a list of optimization problems they worked, which spans a wide variety of sectors, from finance to entertainment, from healthcare to education.
I enjoy a lot hearing about OR projects that are different in terms of the application area or how different and creative the solution methodology is. I urge you to share if you found anything interesting in the program as a comment here or on Twitter with me. If you are like me, then don’t miss Daniel H. Wagner Prize session where my advisor Prof. Terlaky will share their work on inmate assignment and scheduling problem for PA Department of Corrections.