Timezone: »

 
Talk
Manifold Learning
Lawrence Saul

Thu Dec 09 02:30 PM -- 03:00 PM (PST) @

How can we detect low dimensional structure in high dimensional data?  Sam and I worked feverishly on this problem for a number of years.  We were particularly interested in analyzing high dimensional data that lies on or near a low dimensional manifold.  I will describe the algorithm, locally linear embedding (LLE), that we developed for this problem.  I will conclude by relating LLE to more recent work in manifold learning and sketching some future directions for research.

Author Information

Lawrence Saul (Flatiron Institute)

More from the Same Authors