Skip to content

Credits and licensing

Squeeze is maintained by George Pearse. It combines an inherited UMAP implementation with additional Rust dimensionality-reduction methods and benchmark tooling.

UMAP lineage

The UMAP code derives from umap-learn, created by Leland McInnes and collaborators. Rewriting Squeeze's documentation does not remove that authorship or change the source license. See the repository's license for the retained terms and notices.

When publishing work using UMAP, cite the original algorithm and software:

  • McInnes, Healy and Melville, UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction, arXiv:1802.03426.
  • McInnes, Healy, Saul and Grossberger, UMAP: Uniform Manifold Approximation and Projection, JOSS 3(29), 861.

Squeeze's Rust implementations should be distinguished from the original authors' reference implementations. Matching an algorithm name does not imply identical numerics, parameters, behavior, or measured results.

Datasets and dependencies

Digits is loaded through scikit-learn. Fashion-MNIST is provided by Zalando Research; see its repository and citation instructions. Squeeze depends on NumPy, SciPy, scikit-learn, Numba, PyNNDescent, PyO3, ndarray, Rayon and other projects listed in the lockfiles.

For a Squeeze experiment, record the repository commit alongside the dataset, algorithm references and protocol. NeighborMap and SpectralMap are experimental methods in this repository, not established substitutes for a cited UMAP result.