jsDist: Jensen–Shannon distance metric (i.e. the square root of the...

In sccore: Core Utilities for Single-Cell RNA-Seq

Jensen–Shannon distance metric (i.e. the square root of the Jensen–Shannon divergence) between the columns of a dense matrix m

Description

Jensen–Shannon distance metric (i.e. the square root of the Jensen–Shannon divergence) between the columns of a dense matrix m

Usage

jsDist(m, ncores = 1L)

Arguments

m	Input matrix
ncores	Number of threads to be set via omp_set_num_threads() for RcppArmadillo

Value

Vectorized version of the lower triangle as an R distance object, stats::dist()

Examples

ex = matrix(1:9, nrow = 3, ncol = 3)
# JS distance calculated between columns of input matrix
jsDist(ex)


# To demonstrate how the above JS Distance to the JS Divergence, 
# we use the third-party function 'philentropy::JSD()', which 
# computes the JS divergence between rows of the input matrix. 
# The following will give the same results as 'jsDist(ex)':
sqrt(philentropy::JSD(t(ex), est.prob = "empirical"))


# Conversely, we can use the column-normalized matrix, 
# and ignore the argument 'est.prob = "empirical"' from 'philentropy::JSD()', 
# which calculates the relative frequencies of each vector are computed internally). 
# This again will give the same results as 'jsDist(ex)':
ex_cnorm = t( t(ex)/colSums(ex) )
sqrt(philentropy::JSD(t(ex_cnorm)))


# Again obviously 'jsDist(ex)**2' will be the JS divergence, 
# equaling 'philentropy::JSD(t(ex_cnorm))' and 'philentropy::JSD(t(ex), est.prob = "empirical")'
jsDist(ex)**2 
philentropy::JSD(t(ex_cnorm))
philentropy::JSD(t(ex), est.prob = "empirical")

sccore documentation built on April 6, 2026, 5:06 p.m.