Timezone: »

 
A millennium of nearest neighbor methods – an introduction to the NIPS nearest neighbor workshop 2017
George H Chen

Fri Dec 08 09:00 AM -- 09:20 AM (PST) @
in Nearest Neighbors for Modern Applications with Massive Data: An Age-old Solution with New Challenges »

I give a brief history of nearest neighbor (NN) methods, starting from Alhazen’s “Book of Optics” in the 11th century, and leading up to present time. Surprisingly, the most general nonasymptotic results on NN prediction only recently emerged in 2014 in the work of Chaudhuri and Dasgupta. Turning toward “user-friendly” theory with an eye toward practitioners, I mention recent guarantees (2013-2015) in the contemporary applications of time series forecasting, online collaborative filtering, and medical image segmentation — in all three cases, clustering structure enables successful NN prediction.

Author Information

George H Chen (Carnegie Mellon University)

George Chen is an assistant professor of information systems at Carnegie Mellon University. He works on nonparametric prediction methods, applied to healthcare and sustainable development. He received his PhD from MIT in Electrical Engineering and Computer Science.

More from the Same Authors