Two Microsoft researchers elected to National Academy of Sciences
By Allison Linn, Senior Writer & George Thomas, Jr., Writer, Microsoft
Yuval Peres and Robert Schapire, two Microsoft researchers known for their groundbreaking achievements in their respective fields, have been elected to the National Academy of Sciences.
The two researchers are among 84 new members and 21 foreign associates who are joining the National Academy this year. A membership in the National Academy is one of the highest honors in science and research, and it recognizes the researchers’ distinguished and continuing achievements in original research.
Schapire, a principal researcher in Microsoft’s New York City research lab, is a leading thinker in the field of machine learning, in which systems get better at making accurate predictions based on past data they have received. Machine learning algorithms are widely used for everything from sorting spam out of your e-mail inbox to finding faces when you take a picture with a modern digital camera.
Schapire is best known for being one of the inventors of a widely used approach to machine learning called “boosting,” which allows researchers to greatly improve the accuracy of any given machine learning algorithm.
Boosting provides a principled method for creating a highly accurate machine learning model even if you only start with models that are incredibly weak in terms of their predictive power. The idea is to train many of these weak models, and then to combine them together into one very strong model.
Schapire also has been a pioneer and key contributor to many other areas of machine learning research. Most recently, he has been working with other Microsoft researchers on an application for Project Malmo, which uses Minecraft as a platform for generalized artificial intelligence research. The research also is incorporating boosting methods.
Specifically, Peres is well known for his work on ways to bridge the gap between various fields that use probability theory, by looking at situations where multiple fields have comparable problems to solve and can use similar mathematical tools.
For example, at one point he looked at computer science, mathematical biology and statistical physics. He found that biologists struggle to accurately identify the critical mutation rate for genetic data to survive over generations, while computer scientists struggle with how to preserve and replicate information in a way that will be resilient to errors or data loss. Meanwhile, statistical physicists are looking at the effects of temperature on long-range correlations.
It turns out, all of these problems have common characteristics, and the researchers ended up using similar mathematical models in order to study them. By bringing those researchers together, he was able to help them gain insight from one another.
More recently, he’s applied a similar approach to helping his colleagues at Microsoft who are working on theoretical machine learning problems, including trying to understand the effects of switching costs. That’s an area of research that seeks to understand the effect of switching from one choice to another, such as taking a different route to work.
The National Academy of Sciences is a private, nonprofit organization that was established by an act of Congress in 1863. It is charged with providing the nation with scientific advice and furthering the study of science in America. There are currently about 2,250 members and about 440 foreign associates of the academy, and about 200 of those have received Nobel prizes.
Five other Microsoft researchers are members of the National Academy of Sciences.