RunICP: Iterative Clustering Projection (ICP) clustering

View source: R/Utils.R

RunICPR Documentation

Iterative Clustering Projection (ICP) clustering

Description

The function implements Iterative Clustering Projection (ICP): a supervised learning -based clustering, which maximizes clustering similarity between the clustering and its projection by logistic regression.

Usage

RunICP(
  normalized.data = NULL,
  k = 15,
  d = 0.3,
  r = 5,
  C = 5,
  reg.type = "L1",
  max.iter = 200,
  icp.batch.size = Inf
)

Arguments

normalized.data

A sparse matrix (dgCMatrix) containing normalized gene expression data with genes in rows and cells in columns. Default is NULL.

k

A positive integer greater or equal to 2, denoting the number of clusters in ICP. Default is 15.

d

A numeric that defines how many cells per cluster should be down- and oversampled (d in ceiling(N/k*d)), when stratified.downsampling=FALSE, or what fraction should be downsampled in the stratified approach ,stratified.downsampling=TRUE. Default is 0.3.

r

A positive integer that denotes the number of reiterations performed until the algorithm stops. Default is 5.

C

Cost of constraints violation (C) for L1-regulatization. Default is 0.3.

reg.type

"L1" for LASSO and "L2" for Ridge. Default is "L1".

max.iter

A positive integer that denotes the maximum number of iterations performed until the algorithm ends. Default is 200.

icp.batch.size

A positive integer that specifies how many cells to randomly select for each ICP run from the complete data set. This is a new feature intended to speed up the process with larger data sets. Default is Inf, which means using all cells.

Value

A list that includes the probability matrix and the clustering similarity measures: ARI, NMI, etc.


elolab/iloreg documentation built on March 27, 2022, 4:19 a.m.