An S4 Class implementing the UMAP algorithm
Details
Uniform Manifold Approximation is a gradient descend based algorithm that gives results similar to t-SNE, but scales better with the number of points.
Slots
funA function that does the embedding and returns a dimRedResult object.
stdparsThe standard parameters for the function.
General usage
Dimensionality reduction methods are S4 Classes that either be used
directly, in which case they have to be initialized and a full
list with parameters has to be handed to the @fun()
slot, or the method name be passed to the embed function and
parameters can be given to the ..., in which case
missing parameters will be replaced by the ones in the
@stdpars.
Parameters
UMAP can take the follwing parameters:
- ndim
The number of embedding dimensions.
- knn
The number of neighbors to be used.
- d
The distance metric to use.
- method
"naive"for an R implementation,"python"for the reference implementation.
Other method parameters can also be passed, see
umap.defaults for details. The ones above have been
standardized for the use with dimRed and will get automatically
translated for umap.
Implementation
The dimRed package wraps the umap packages which provides
an implementation in pure R and also a wrapper around the original python
package umap-learn (https://github.com/lmcinnes/umap/). This requires
umap-learn version 0.4 installed, at the time of writing, there is
already umap-learn 0.5 but it is not supported by the R package
umap.
The "naive" implementation is a pure R implementation and considered
experimental at the point of writing this, it is also much slower than the
python implementation.
The "python" implementation is the reference implementation used by
McInees et. al. (2018). It requires the reticulate
package for the interaction with python and the python package
umap-learn installed (use pip install umap-learn).
References
McInnes, Leland, and John Healy. "UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction." https://arxiv.org/abs/1802.03426
See also
Other dimensionality reduction methods:
AutoEncoder-class,
DRR-class,
DiffusionMaps-class,
DrL-class,
FastICA-class,
FruchtermanReingold-class,
HLLE-class,
Isomap-class,
KamadaKawai-class,
MDS-class,
NNMF-class,
PCA-class,
PCA_L1-class,
dimRedMethod-class,
dimRedMethodList(),
kPCA-class,
nMDS-class,
tSNE-class
Examples
if (FALSE) {
dat <- loadDataSet("3D S Curve", n = 300)
emb <- embed(dat, "UMAP", .mute = NULL, knn = 10)
plot(emb, type = "2vars")
}