Type: Package
Title: Dimensionality Reduction Methods in a Common Format
Version: 1.0.4
Description: Provides a common interface for applying dimensionality reduction methods, such as Principal Component Analysis ('PCA'), Independent Component Analysis ('ICA'), diffusion maps, Locally-Linear Embedding ('LLE'), t-distributed Stochastic Neighbor Embedding ('t-SNE'), and Uniform Manifold Approximation and Projection ('UMAP'). Has built-in support for sparse matrices.
License: MIT + file LICENSE
Encoding: UTF-8
LazyData: true
Imports: dynutils (≥ 1.0.5), irlba, lmds, stats, tibble
Suggests: testthat, diffusionMap, fastICA, igraph, lle, MASS, Matrix, RANN, Rtsne, smacof, uwot
RoxygenNote: 7.1.1
URL: https://github.com/dynverse/dyndimred
BugReports: https://github.com/dynverse/dyndimred/issues
NeedsCompilation: no
Packaged: 2021-03-23 07:53:59 UTC; rcannood
Author: Robrecht Cannoodt ORCID iD [aut, cre] (github: rcannood), Wouter Saelens ORCID iD [aut] (github: zouter)
Maintainer: Robrecht Cannoodt <rcannood@gmail.com>
Repository: CRAN
Date/Publication: 2021-03-23 08:30:06 UTC

Perform simple dimensionality reduction

Description

Perform simple dimensionality reduction

Usage

dimred(x, method, ndim, ...)

dimred_dm_destiny(
  x,
  ndim = 2,
  distance_method = c("euclidean", "spearman", "cosine")
)

dimred_dm_diffusionmap(
  x,
  ndim = 2,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski")
)

dimred_ica(x, ndim = 3)

dimred_knn_fr(
  x,
  ndim = 2,
  lmds_components = 10,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski"),
  n_neighbors = 10
)

dimred_landmark_mds(
  x,
  ndim = 2,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski")
)

dimred_lle(x, ndim = 3)

dimred_mds(
  x,
  ndim = 2,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski")
)

dimred_mds_isomds(
  x,
  ndim = 2,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski")
)

dimred_mds_sammon(
  x,
  ndim = 2,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski")
)

dimred_mds_smacof(
  x,
  ndim = 2,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski")
)

dimred_pca(x, ndim = 2)

list_dimred_methods()

Arguments

x

Log transformed expression data, with rows as cells and columns as features

method

The name of the dimensionality reduction method to use

ndim

The number of dimensions

...

Any arguments to be passed to the dimensionality reduction method

distance_method

The name of the distance metric, see dynutils::calculate_distance

lmds_components

The number of lmds components to use. If NULL, LMDS will not be performed first. If this is a matrix, it is assumed it is a dimred for x.

n_neighbors

The size of local neighborhood (in terms of number of neighboring sample points).

Examples

library(Matrix)
x <- abs(Matrix::rsparsematrix(100, 100, .5))
dimred(x, "pca", ndim = 3)
dimred(x, "ica", ndim = 3)

if (interactive()) {
  dimred_dm_destiny(x)
  dimred_dm_diffusionmap(x)
  dimred_ica(x)
  dimred_landmark_mds(x)
  dimred_lle(x)
  dimred_mds(x)
  dimred_mds_isomds(x)
  dimred_mds_sammon(x)
  dimred_mds_smacof(x)
  dimred_pca(x)
  dimred_tsne(x)
  dimred_umap(x)
}

tSNE

Description

tSNE

Usage

dimred_tsne(
  x,
  ndim = 2,
  perplexity = 30,
  theta = 0.5,
  initial_dims = 50,
  distance_method = c("pearson", "spearman", "cosine", "euclidean", "chisquared",
    "hamming", "kullback", "manhattan", "maximum", "canberra", "minkowski")
)

Arguments

x

Log transformed expression data, with rows as cells and columns as features

ndim

The number of dimensions

perplexity

numeric; Perplexity parameter (should not be bigger than 3 * perplexity < nrow(X) - 1, see details for interpretation)

theta

numeric; Speed/accuracy trade-off (increase for less accuracy), set to 0.0 for exact TSNE (default: 0.5)

initial_dims

integer; the number of dimensions that should be retained in the initial PCA step (default: 50)

distance_method

The name of the distance metric, see dynutils::calculate_distance

See Also

Rtsne::Rtsne()

Examples

library(Matrix)
dataset <- abs(Matrix::rsparsematrix(100, 100, .5))
dimred_tsne(dataset, ndim = 3)

UMAP

Description

UMAP

Usage

dimred_umap(
  x,
  ndim = 2,
  distance_method = c("euclidean", "cosine", "manhattan"),
  pca_components = 50,
  n_neighbors = 15L,
  init = "spectral",
  n_threads = 1
)

Arguments

x

Log transformed expression data, with rows as cells and columns as features

ndim

The number of dimensions

distance_method

The name of the distance metric, see dynutils::calculate_distance

pca_components

The number of pca components to use for UMAP. If NULL, PCA will not be performed first

n_neighbors

The size of local neighborhood (in terms of number of neighboring sample points).

init

Type of initialization for the coordinates. Options are:

  • "spectral" Spectral embedding using the normalized Laplacian of the fuzzy 1-skeleton, with Gaussian noise added.

  • "normlaplacian". Spectral embedding using the normalized Laplacian of the fuzzy 1-skeleton, without noise.

  • "random". Coordinates assigned using a uniform random distribution between -10 and 10.

  • "lvrandom". Coordinates assigned using a Gaussian distribution with standard deviation 1e-4, as used in LargeVis (Tang et al., 2016) and t-SNE.

  • "laplacian". Spectral embedding using the Laplacian Eigenmap (Belkin and Niyogi, 2002).

  • "pca". The first two principal components from PCA of X if X is a data frame, and from a 2-dimensional classical MDS if X is of class "dist".

  • "spca". Like "pca", but each dimension is then scaled so the standard deviation is 1e-4, to give a distribution similar to that used in t-SNE. This is an alias for init = "pca", init_sdev = 1e-4.

  • "agspectral" An "approximate global" modification of "spectral" which all edges in the graph to a value of 1, and then sets a random number of edges (negative_sample_rate edges per vertex) to 0.1, to approximate the effect of non-local affinities.

  • A matrix of initial coordinates.

For spectral initializations, ("spectral", "normlaplacian", "laplacian"), if more than one connected component is identified, each connected component is initialized separately and the results are merged. If verbose = TRUE the number of connected components are logged to the console. The existence of multiple connected components implies that a global view of the data cannot be attained with this initialization. Either a PCA-based initialization or increasing the value of n_neighbors may be more appropriate.

n_threads

Number of threads to use (except during stochastic gradient descent). Default is half the number of concurrent threads supported by the system. For nearest neighbor search, only applies if nn_method = "annoy". If n_threads > 1, then the Annoy index will be temporarily written to disk in the location determined by tempfile.

See Also

uwot::umap()

Examples

library(Matrix)
dataset <- abs(Matrix::rsparsematrix(100, 100, .5))
dimred_umap(dataset, ndim = 2, pca_components = NULL)

Common dimensionality reduction methods

Description

Provides a common interface for applying common dimensionality reduction methods, Such as PCA, ICA, diffusion maps, LLE, t-SNE, and umap.