Microsoft Research

Microsoft Research

The Contextual Bandits Problem: A Fast, Simple, and Optimal Algorithm 2017

Thursday, May 4, 2017 (All day)
Columbia University
New York, NY 10027
United States

Colloquium Series: Dr. Robert Schapire
Principal Researcher, Microsoft Research (NYC)

We study the general problem of how to learn through experience to make intelligent decisions.  In this setting, called the contextual bandits problem, the learner must repeatedly decide which action to take in response to an observed context, and is then permitted to observe the received reward, but only for the chosen action.  The goal is to learn through experience to behave nearly as well as the best policy (or decision rule) in some possibly very large and rich space of candidate policies.  Previous approaches to this problem were all highly inefficient and often extremely complicated.  In this work, we present a fast and simple algorithm that learns to behave as well as the best policy at a rate that is (almost) statistically optimal.  Our approach assumes access to a kind of “oracle” (or subroutine) for classification learning problems which can be used to select policies; in practice, most off-the-shelf classification algorithms could be used for this purpose.  Our algorithm makes very modest use of the oracle, which it calls far less than once per round, on average, a huge improvement over previous methods.  These properties suggest this may be the most practical contextual bandits algorithm among all existing approaches that are provably effective for general policy classes.

This is joint work with Alekh Agarwal, Daniel Hsu, Satyen Kale, John Langford and Lihong Li. 

Once Upon a Graph: How to Get from Now to Then in Massive Networks

Thursday, April 20, 2017 - 5:00pm to 6:00pm
Davis Auditorium | The Schapiro Center
Columbia University
New York, NY 10027
United States

The Distinguished Colloquium Series in Interdisciplinary and Applied Mathematics, along with Columbia's Data Science Institute, proudly present a lecture by:

Jennifer Tour Chayes
Distinguished Scientist and Managing Director of Microsoft Research New England, Cambridge, MA.

Title: "Once upon a graph: How to get from now to then in massive networks"

Jennifer Tour Chayes

Jennifer Tour Chayes
Microsoft Research New England, Cambridge, MA
Distinguished Scientist
Managing Director

Jennifer Tour Chayes is Distinguished Scientist and Managing Director of Microsoft Research New England in Cambridge, Massachusetts, which she co-founded in 2008, and Microsoft Research New York City, which she co-founded in 2012. Chayes was Research Area Manager for Mathematics, Theoretical Computer Science and Cryptography at Microsoft Research Redmond until 2008. Chayes joined Microsoft Research in 1997, when she co-founded the Theory Group.

Crashing Drones and Hijacked Cameras: CyberTrust Meets CyberPhysical

Monday, May 9, 2016 - 1:30pm to 2:30pm
Costa Engineering Commons, 750 CEPSR
530 West 120th Street, 7th Floor
New York, NY 10027
United States

Cyber-physical systems are engineered systems that require tight conjoining of and coordination between the computational (discrete) and the physical (continuous). Cyber-physical systems are rapidly penetrating every aspect of our lives, with potential impact on sectors critical to national security and competitiveness, including aerospace, automotive, chemical production, civil infrastructure, energy, finance, healthcare, manufacturing, materials, and transportation. As these systems fulfill the promise of the Internet of Things, smart cities, household robots, and personalized medicine, we need to ensure they are trustworthy: reliable, secure, and privacy-preserving. This talk will look at cyber-physical systems from the lens of trustworthy computing. Throughout my talk, I will raise research challenges for how to make cyber-physical systems trustworthy, with a special emphasis on privacy.

Colloquium Series: Jeannette Wing
Corporate Vice President, Microsoft Research

Jeannette Wing

Jeannette Wing
Microsoft Research
Corporate Vice President

Jeannette M. Wing is Corporate Vice President, Microsoft Research. She is Adjunct Professor of Computer Science at Carnegie Mellon where she twice served as the Head of the Computer Science Department. From 2007-2010 she was the Assistant Director of the Computer and Information Science and Engineering Directorate at the National Science Foundation. She received her S.B., S.M., and Ph.D. degrees in Computer Science, all from the Massachusetts Institute of Technology.

Robert Schapire

Microsoft Research
Principal Researcher

Robert Schapire is a Principal Researcher at Microsoft Research in New York City. He received his PhD from MIT in 1991. After a short post-doc at Harvard, he joined the technical staff at AT&T Labs (formerly AT&T Bell Laboratories) in 1991. In 2002, he became a Professor of Computer Science at Princeton University. He joined Microsoft Research in 2014. His awards include the 1991 ACM Doctoral Dissertation Award, the 2003 Gödel Prize, and the 2004 Kanelakkis Theory and Practice Award (both of the last two with Yoav Freund).

Once upon a graph: How to get from now to then in massive networks

Wednesday, April 20, 2016 - 1:00pm to 2:00pm
United States

Networks are increasingly important in many aspects of our world: physical networks like transportation networks, utility networks and the Internet, online information networks like the WWW, online social networks like Facebook and Twitter, epidemiological networks for global disease transmission, genomic and protein networks in computational biology, and many more. How do we model and learn these networks? In contrast to conventional learning problems, where we have many independent samples, it is often the case for these networks that we can get only one independent sample.

Jake Hofman

Jake Hofman

Jake Hofman is a Senior Researcher at Microsoft Research in New York City, where his work in computational social science involves applications of statistics and machine learning to large-scale social data. Prior to joining Microsoft, he was a member of the Microeconomics and Social Systems group at Yahoo! Research. Jake is also an Adjunct Assistant Professor of Applied Mathematics at Columbia University, where he has designed and taught classes on a number of topics ranging from biological physics to applied machine learning. He holds a B.S.

Pages

Back to Top