Timezone: »

 
Poster
Improved Graph Laplacian via Geometric Self-Consistency
Dominique Perrault-Joncas · Marina Meila · James McQueen

Mon Dec 04 06:30 PM -- 10:30 PM (PST) @ Pacific Ballroom #46
in Posters Mon »

We address the problem of setting the kernel bandwidth, epps, used by Manifold Learning algorithms to construct the graph Laplacian. Exploiting the connection between manifold geometry, represented by the Riemannian metric, and the Laplace-Beltrami operator, we set epps by optimizing the Laplacian's ability to preserve the geometry of the data. Experiments show that this principled approach is effective and robust

Keywords: Kernel Methods · Semi-Supervised Learning · Unsupervised Learning · Nonlinear Dimensionality Reduction and Manifold Learning · Hyperparameter Selection
[ Paper] 

Author Information

Dominique Perrault-Joncas (Google)
Marina Meila (University of Washington)
James McQueen (University of Washington)

More from the Same Authors