tauInitial function

List of initial values for $\tau$

Same function whatever directed or undirected case

tauInitial(data, n, Q, d_part, n_perturb, perc_perturb, n_random, directed)

Arguments

• data: Data : only needs the $N_{ijk}$ field of data

• n: Total number of nodes

• Q: Total number of groups

• d_part: Maximal level for finest partitions of time interval [0,T], used for kmeans initializations.

  • Algorithm takes partition up to depth $2^d$ with $d=1,...,d_{part}$
  • Explore partitions $[0,T], [0,T/2], [T/2,T], ... [0,T/2^d], ...[(2^d-1)T/2^d,T]$
  • Total number of partitions $npart= 2^{(d_part +1)} - 1$

• n_perturb: Number of different perturbations on k-means result

• perc_perturb: Percentage of labels that are to be perturbed (= randomly switched)

• n_random: Number of completely random initial points. If not zero there will be n_random taus uniformly sampled in the initialization.

• directed: Boolean for directed (TRUE) or undirected (FALSE) case

Returns

List of matrixes of initial values for $\tau$

Details

The (maximal) total number of initializations is $d_{part}*(1+n_{perturb}) + n_{random}$

Examples

# Generate initial tau for generated_Q3 data

n <- 50
Dmax <- 2^3
Q <- 3
d_part <- 1 # less than 3 (owing to Dmax)
n_perturb <- 2
perc_perturb <- 0.2
n_random <- 1
directed <- FALSE

data <- list(Nijk = statistics(generated_Q3$data, n, Dmax, directed = FALSE))

tau <- tauInitial(data,n,Q,d_part,n_perturb,perc_perturb,n_random,directed)