PrototypesCriticisms

importances(observed_ranking, rankings)

Compute the importances for every data instance.

Arguments

• observed_ranking::Vector{Int}: the originally observed ranking.
• rankings::Matrix{Int}: the rankings per data feature.

importances(observed_ranking, ranking)

Like importances, but provide a single ranking for one data feature only.

importances(observed_ranking, rankings)

Like importances, but provide a vector of rankings instead of a matrix.

prototypes_criticisms(observed_ranking::Vector{Int}, D::Matrix, k::Int)

Find prototypes and criticisms based on the clustering of the ranking.

Arguments

• observed_ranking::Vector{Int}: the observed ranking.
• D::AbstractMatrix: the data instances.
• k::Int: the number of clusters to find.
• metric::Distances.PreMetric: the distance metric for computing the distances between the data instances.

Like prototypes_criticisms, but provide a pre-computed distance matrix instead of a matrix of data instances and a distance metric.

prototypes_criticisms(observed_ranking, rankings, p, c)

Find prototypes and criticisms based on the data feature importances.

Arguments

• observed_ranking::Vector{Int}: the observed ranking.
• rankings::Matrix{Int}: the rankings per data feature.
• p::Int: the number of protoypes to find.
• c::Int: the number of criticisms to find.

Notes

Each ranking contains the data instance identifiers, i.e., every ranking is a mapping of indices to data instances, e.g., observed_ranking[3] = "data instance at rank 3"

prototypes_criticisms(observed_ranking, rankings, p, c)

Like prototypes_criticisms, but provide a vector of rankings instead of a matrix.