Compute the KL Divergence

From Wikipedia: "The Kullback–Leibler (KL) divergence (also called relative entropy and I-divergence) is a type of statistical distance that measures how one probability distribution P is different from a second, reference probability distribution Q. A simple interpretation of the KL divergence of P from Q is the expected excess surprise from using Q as a model when the actual distribution is P." In scrattch.mapping this distance is required for computing categorical confidence calls used for assessing cell quality.

compute_KLdiv(
  query_probabilities,
  reference_probabilities,
  select.cl = NULL,
  select.cells = NULL
)

Arguments

query_probabilities

A matrix of query cell mapping probabilities where rows represents cells, columns represent clusters and values rowsum to 1.

reference_probabilities

A matrix of reference cluster probabilities where rows and columns both represents clusters values represent confusion between cluster mappings/clusterings and rowsum to 1.

select.cl

An (optional) vector of cluster ids (e.g., row/col names of reference_probabilities) representing clusters for which KL divergence should be calculated using

select.cells

An (optional) vector of cell ids (e.g., row names of query_probabilities) representing cells for which KL divergence should be calculated on

Value

A matrix of KL divergences for each requested cell (row) in each requested cluster (column)