dTBM (version 3.0)

dtbm: Multiway spherical clustering for degree-corrected tensor block model

Description

Multiway spherical clustering for degree-corrected tensor block model including weighted higher-order initialization and angle-based iteration. Main function in the package. This function takes the tensor/matrix observation, the cluster number, and a logic variable indicating the symmetry as input. Output contains initial and refined clustering assignment.

Usage

dtbm(Y, r, max_iter, alpha1 = 0.01, asymm)

Value

a list containing the following:

z

a list of vectors recording the refined clustering assignment with initialization z0

s_deg

logic variable, if "TRUE", degenerate estimated core tensor/matrix occurs during the iteration; if "FALSE", otherwise

z0

a list of vectors recording the initial clustering assignment

s0

a list of vectors recording the index of degenerate entities with random clustering assignment in initialization

Arguments

Y

array/matrix, order-3 tensor/matrix observation

r

vector, the cluster number on each mode; see "details"

max_iter

integer, max number of iterations if update does not converge

alpha1

number, substitution of degenerate core tensor; see "details"

asymm

logic variable, if "TRUE", assume the clustering assignment differs in different modes; if "FALSE", assume all the modes share the same clustering assignment

Details

r should be a length 2 vector for matrix and length 3 vector for tensor observation;

all the elements in r should be integer larger than 1;

symmetric case only allow r with the same cluster number on each mode;

observations with non-identical dimension on each mode are only applicable with asymm = T.

When the estimated core tensor has a degenerate slice during iteration, i.e., a slice with all zero elements, randomly pick an entry in the degenerate slice with value alpha1.

Examples

test_data = sim_dTBM(seed = 1, imat = FALSE, asymm = FALSE, p = c(50,50,50), r = c(3,3,3),
                    core_control = "control", s_min = 0.05, s_max = 1,
                    dist = "normal", sigma = 0.5,
                    theta_dist = "pareto", alpha = 4, beta = 3/4)

result = dtbm(test_data$Y, r = c(3,3,3), max_iter = 20, asymm = FALSE)