Comparison To Other Methods
Semidefinite embedding is much better in revealing the underlying dimension of the data compared to LLE and Laplacian eigenmaps. It also guarantees that the nearest neighbors in the embedding is the same as the original nearest neighbor for each point while the other two methods do not. On the other hand, semidefinite embedding is much slower and harder to scale to large data.
Semidefinite embedding outperforms Isomap when the manifold is not a convex subset of the Euclidean space.
Read more about this topic: Semidefinite Embedding
Famous quotes containing the words comparison and/or methods:
“[Girls] study under the paralyzing idea that their acquirements cannot be brought into practical use. They may subserve the purposes of promoting individual domestic pleasure and social enjoyment in conversation, but what are they in comparison with the grand stimulation of independence and self- reliance, of the capability of contributing to the comfort and happiness of those whom they love as their own souls?”
—Sarah M. Grimke (1792–1873)
“Cold and hunger seem more friendly to my nature than those methods which men have adopted and advise to ward them off.”
—Henry David Thoreau (1817–1862)