chooseClusteringSize {tensorsparse}

Select the clustering size for order-3 sparse tensor clustering via BIC

Description

Select the clustering size for three-way clustering. The function searches over a range of clustering sizes and outputs the one that minimizes BIC. The clustering size (d_1, d_2, d_3) is a length-3 vector consisting of the number of clusters in each mode.

Usage

chooseClusteringSize(
  x,
  k,
  r,
  l,
  lambda = 0,
  sim.times = 1,
  method = "L0",
  n.cores = NULL
)

Arguments

x

a three-dimensional array

k

a vector, the possible numbers of clusters at mode 1

r

a vector, the possible numbers of clusters at mode 2

l

a vector, the possible numbers of clusters at mode 3

lambda

a numeric value, regularization coefficient

sim.times

the number of simulation replicates when performing clustering

method

two options: "L0", "L1". "L0" indicates L0 penalty, and "L1" indicates Lasso penalty

n.cores

the number of cores in parallel implementation

Value

a list

estimated_krl a 1*3 matrix consisting of the estimated clustering size

BIC a vector consisting of the BIC value for all combinations of clustering sizes