Expand description
Manifold learning algorithms.
Provides:
-
Classical MDS — eigendecomposition variant of
sklearn.manifold.MDS. -
Isomap —
sklearn.manifold.Isomap: k-NN graph + geodesic shortest paths (Floyd-Warshall) + classical MDS on the geodesic distances. -
LocallyLinearEmbedding —
sklearn.manifold.LocallyLinearEmbedding: local reconstruction weights + bottom-k eigenvectors of(I − W)ᵀ(I − W). -
t-SNE —
sklearn.manifold.TSNE: vanilla O(n²) gradient descent with perplexity-calibrated Gaussian affinities and student-t low-dim affinities. Suitable for n ≲ 1000.
Re-exports§
pub use isomap::FittedIsomap;pub use isomap::Isomap;pub use lle::FittedLocallyLinearEmbedding;pub use lle::LocallyLinearEmbedding;pub use tsne::FittedTSne;pub use tsne::TSne;pub use tsne::TSneMethod;
Modules§
- isomap
- Isomap — non-linear dimensionality reduction via geodesic distances.
- lle
- Locally Linear Embedding (LLE).
- tsne
- t-distributed Stochastic Neighbor Embedding.