HSIC: Hilbert-Schmidt Independence Criteria (HSIC)

In rikenbit/fuchikoma: Revealing differentially expressed genes using nonlinear dimensionality reduction for single cell RNA-Seq data

Description

Inner function of fuchikoma. HSIC calculated aginst two gram-matrix means the independence of two kernel spaces. The higher HSIC value is, the more two space are dependent. When p.value set TRUE, p-value of HSIC is also calculated by moment matching to a gamma distribution.

Usage

HSIC(K, L, shrink=FALSE, type=c("gamma", "permutation"), n.perm=100)

Arguments

K	First gram matrix
L	Second gram matrix
shrink	Shrinkage option when the sample size is insufficient (default:FALSE)
type	When type is gamma, p-value is calculated under the assumption of gamma distribution. When type is permutation, p-value is calculated by random permutation test (default:gamma)
n.perm	Number of permutation (default:100)

Value

HSIC	The value of HSIC
Pval	P-value of HSIC. The value is accessible only when p.value is specified as TRUE

Author(s)

Koki Tsuyuzaki, Haruka Ozaki, Mika Yoshimura, Itoshi Nikaido

Maintainer: Koki Tsuyuzaki <[e-mail]>

References

Le Song et al. (2007) Gene selection via the BAHSIC family of algorithms, Bioinformatics, 23(13), i490-i498

Y-h Taguchi et al. (2015) Principal component analysis-based unsupervised feature extraction applied to in silico drug discovery for posttraumatic stress disorder-mediated heart disease, BMC Bioinformatics, 16(139)

Arthur Gretton et al. (2007) A Kernel Statistical Test of Independence, NIPS 21

Aaditya Ramdas et al. (2015) Nonparametric Independence Testing for Small Sample Sizes, IJCAI-15

Examples

K <- matrix(runif(100), nrow=10)
L <- matrix(runif(100), nrow=10)
HSIC(K, L)

rikenbit/fuchikoma documentation built on May 27, 2019, 9:09 a.m.