Compute co-ranking quality and behavior indices

Given high-dimensional and low-dimensional distance matrices, computes quality (Q) and behavior (B) indices based on the co-ranking matrix framework. Three index families are computed: Trustworthiness & Continuity (T&C), Mean Relative Rank Errors (MRRE), and Local Continuity Meta-Criterion (LCMC).

Usage

coranking_indices(Dh, Dl, K)

Arguments

Dh

Numeric matrix of high-dimensional pairwise distances (n x n).

Dl

Numeric matrix of low-dimensional pairwise distances (n x n).

K

Integer neighborhood size.

Value

A named numeric vector with elements Q_TC, B_TC, Q_RE, B_RE, Q_LC, B_LC.

Examples

set.seed(42)
X <- matrix(rnorm(50), 10, 5)
Y <- matrix(rnorm(20), 10, 2)
Dh <- as.matrix(dist(X))
Dl <- as.matrix(dist(Y))
coranking_indices(Dh, Dl, K = 3)
#>        Q_TC        B_TC        Q_RE        B_RE        Q_LC        B_LC 
#> 0.653333333 0.066666667 0.670000000 0.008235294 0.500000000 0.266666667