| NCV.select {randnet} | R Documentation |
selecting block models by NCV
Description
selecting block models by NCV of Chen and Lei (2016)
Usage
NCV.select(A, max.K, cv = 3)
Arguments
A |
adjacency matrix |
max.K |
largest number of communities to check |
cv |
fold of cross-validation |
Details
Spectral clustering is used for fitting the block models
Value
a list of
dev |
the binomial deviance loss under SBM for each K |
l2 |
the L_2 loss under SBM for each K |
dc.dev |
the binomial deviance loss under DCSBM for each K |
dc.l2 |
the L_2 loss under DCSBM for each K |
dev.model |
the selected model by deviance loss |
l2.model |
the selected model by L_2 loss |
sbm.l2.mat, sbm.dev.mat, .... |
the corresponding matrices of loss for each fold (row) and each K value (column) |
Author(s)
Tianxi Li, Elizaveta Levina, Ji Zhu
Maintainer: Tianxi Li tianxili@virginia.edu
References
Chen, K. & Lei, J. Network cross-validation for determining the number of communities in network data Journal of the American Statistical Association, Taylor & Francis, 2018, 113, 241-251
See Also
Examples
dt <- BlockModel.Gen(30,300,K=3,beta=0.2,rho=0.9,simple=FALSE,power=TRUE)
A <- dt$A
ncv <- NCV.select(A,6,3)
ncv$l2.model
ncv$dev.model
which.min(ncv$dev)
which.min(ncv$l2)
which.min(ncv$dc.dev)
which.min(ncv$dc.l2)