chooseLambda {tensorsparse}

Select the regularization coefficient for order-3 sparse tensor clustering via BIC

Description

Select the regularization coefficient for three-way clustering. The clustering size is assumed to be known. The function searches over a range of regularization sizes and outputs the one that minimizes the BIC.

Usage

chooseLambda(x, k, r, l, lambda = NULL, method = "L0")

Arguments

x

a three-dimensional array

k

an positive integer, the numbers of clusters at mode 1

r

an positive integer, the numbers of clusters at mode 2

l

an positive integer, the numbers of clusters at mode 3

lambda

a vector of possible lambda, eg: lambda = c(0,50,100,200)

method

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

Value

a list

lambda the lambda with lowest BIC

BIC the BIC for each lambda in the given range

nonzeromus the number of clusters with non-zero means