umap v0.2.2.0

Monthly downloads

by Tomasz Konopka View Source

Uniform Manifold Approximation and Projection

Uniform manifold approximation and projection is a technique for dimension reduction. The algorithm was described by McInnes and Healy (2018) in <arXiv:1802.03426>. This package provides an interface for two implementations. One is written from scratch, including components for nearest-neighbor search and for embedding. The second implementation is a wrapper for 'python' package 'umap-learn' (requires separate installation, see vignette for more details).

Functions in umap

Name	Description
identity.coo	Construct an identity matrix
detect.umap.learn	adjust config depending on umap-learn version
predict.umap	project data points onto an existing umap embedding
print.umap.knn	Display summary of knn.info
optimize_epoch	run one epoch of the umap optimization
reduce.coo	Remove some entires in a coo matrix where values are zero
clip	Force (clip) a value into a finite range
check.learn.available	check whether python module is available, abort if not
dManhattan	compute Manhattan distance between two vectors
dEuclidean	compute Euclidean distance between two vectors
find.ab.params	Estimate a/b parameters
knn.from.dist	get information about k nearest neighbors from a distance object or from a matrix with distances
make.initial.spectator.embedding	Create an initial embedding for a set of spectators
dCenteredPearson	compute pearson correlation distance between two vectors
make.epochs.per.sample	Compute a value to capture how often each item contributes to layout optimization
dCosine	compute cosine dissimilarity between two vectors
make.initial.embedding	Create an initial embedding for a graph
knn.info	Compute knn information
mdManhattan	compute Manhattan distances
message.w.date	Send a message() with a prefix with a data
clip4	perform a compound transformation on a vector, including clipping
umap.learn	Create a umap embedding using python package umap-learn
mdCosine	compute cosine distances
concomp.coo	Count the number of connected components in a coo graph
knn.from.data	get information about approximate k nearest neighbors from a data matrix
naive.optimize.embedding	modify an existing embedding
make.random.embedding	Make an initial embedding with random coordinates
knn.from.data.reps	Repeat knn.from.data multiple times, pick the best neighbors
make.spectral.embedding	Create a spectral embedding for a connectivity graph
naive.simplicial.set.embedding	create an embedding of graph into a low-dimensional space
mdEuclidean	compute Euclidean distances
umap.learn.predict	predict embedding of new data given an existing umap object
umap.prep.input	Prep primary input as a data matrix
mdCenteredPearson	compute pearson correlation distances
stop.coo	Stop execution with a custom message
umap.small	Create an embedding object compatible with package umap for very small inputs
multiply.coo	Multiply two coo objects element-wise
umap.naive	Create a umap embedding
umap.naive.predict	predict embedding of new data given an existing umap object
subset.coo	Subset a coo
set.global.seed	set .Random.seed to a pre-saved value
print.umap	Display a summary of a umap object
print.umap.config	Display contents of a umap configuration
smooth.knn.dist	compute a "smooth" distance to the kth neighbor and approximate first neighbor
umap.check.config	Validator functions for umap settings
umap.defaults	Default configuration for umap
umap.error	stop execution with a custom error message
umap.check.config.class	Validator for config class component
umap.warning	create a warning message
naive.fuzzy.simplicial.set	create a simplicial set from a distance object
spectator.knn.info	compute knn information for spectators relative to data
spectral.eigenvectors	get a set of k eigenvectors for the laplacian of x
t.coo	Transpose a coo matrix
umap	Computes a manifold approximation and projection
coo	Create a coo representation of a square matrix
check.coo	Check class for coo
check.compatible.coo	Check that two coo objects are compatible for addition, multiplication
add.coo	Add two coo objects element-wise
coo2mat	Convert from coo object into conventional matrix
center.embedding	Adjust a matrix so that each column is centered around zero
laplacian.coo	Construct a normalized Laplacian for a graph
make.coo	Helper to construct coo objects
get.global.seed	lookup .Random.seed in global environment
No Results!

Vignettes of umap

Name
umap.Rmd
umap_learn.Rmd
No Results!

Last month downloads

Details

License	MIT + file LICENSE
URL	https://github.com/tkonopka/umap
LinkingTo	Rcpp
LazyData	true
VignetteBuilder	knitr
RoxygenNote	6.1.1
NeedsCompilation	yes
Packaged	2019-05-13 05:38:11 UTC; tkonopka
Repository	CRAN
Date/Publication	2019-05-13 06:20:03 UTC

suggests	knitr , rmarkdown , testthat
imports	methods , Rcpp (>= 0.12.6) , reticulate , RSpectra , stats
depends	R (>= 3.1.2)
Contributors