New Algorithms for Manufacturing and Collaborative Filtering

EECS Joint Colloquium Distinguished Lecture Series

Ken Goldberg
Industrial Engineering & Operations Research Center (IEOR), UC Berkeley

Wednesday, November 17, 1999
Hewlett Packard Auditorium, 306 Soda Hall
4:00-5:00 p.m.

Abstract:

The first part of this talk addresses robotics for automated manufacturing. I'll show that any polygonal part can be oriented without sensors using a robot gripper. The proof is based on an algorithm that finds such a plan in time $O(n^2)$ for an $n$-sided planar part. I'll describe how this time complexity has been further reduced and summarize new results in feeding and gripping.

The second part of the talk addresses the very different subject of collaborative filtering, where items are recommended to online users based on their ratings of other items. I'll describe EigenTaste, a new algorithm based on principal component analysis (PCA) that computes predictions in constant time. We are testing the algorithm with Jester, a site that recommends jokes to users based on their ratings of sample jokes. We are also applying the algorithm to investor profiling.

Ken Goldberg is Associate Professor of Industrial Engineering at UC Berkeley. He received his PhD in 1990 from the School of Computer Science at Carnegie Mellon University. Goldberg was named a National Science Foundation Young Investigator in 1994 and NSF Presidential Faculty Fellow in 1995: http://www.ieor.berkeley.edu/~goldberg

231cory@EECS.Berkeley.EDU