S V N Vishwanathan, K. M. Borgwardt, Omri Guttman and Alex Smola
We present a framework for efficient extrapolation of reduced rank approximations,
graph kernels, and locally linear embeddings (LLE) to unseen data. We also present
a principled method to combine many of these kernels and then extrapolate them.
Central to our method is a theorem for matrix approximation, and an extension of
the representer theorem to handle multiple joint regularization constraints. Exper-
iments in protein classification demonstrate the feasibility of our approach.
|Project Keyword:||Project Keyword UNSPECIFIED|
|Subjects:||Theory & Algorithms|
|Deposited By:||Alex Smola|
|Deposited On:||16 January 2006|