Title: Topological methods in dimensionality reduction.

 

Vin de Silva
Stanford University, Mathematics Department

 

Abstract:

There are now many different techniques for finding a low-dimensional representation of a high-dimensional data set. The representation often takes the form of an embedding of the data in low-dimensional Euclidean space. However, it is easy to envisage situations where the true structure of the data may be obscured and not clarified by the embedding; in particular when the low-dimensional structure has non-trivial topology. Motivated by these examples, I discuss some alternative approaches to interpreting high-dimensional data with low-dimensional structure.