objective {corehunter}

Create Core Hunter objective.

Description

The following optimization objectives are supported by Core Hunter:

EN

Average entry-to-nearest-entry distance (default). Maximizes the average distance between each selected individual and the closest other selected item in the core. Favors diverse cores in which each individual is sufficiently different from the most similar other selected item (low redundancy). Multiple distance measures are provided to be used with this objective (see below).

AN

Average accession-to-nearest-entry distance. Minimizes the average distance between each individual (from the full dataset) and the closest selected item in the core (which can be the individual itself). Favors representative cores in which all items from the original dataset are represented by similar individuals in the selected subset. Multiple distance measures are provided to be used with this objective (see below).

EE

Average entry-to-entry distance. Maximizes the average distance between each pair of selected individuals in the core. This objective is related to the entry-to-nearest-entry (EN) distance but less effectively avoids redundant, similar individuals in the core. In general, use of EN is preferred. Multiple distance measures are provided to be used with this objective (see below).

SH

Shannon's allelic diversity index. Maximizes the entropy, as used in information theory, of the selected core. Independently takes into account all allele frequencies, regardless of the locus (marker) where to which the allele belongs. Requires genotypes.

HE

Expected proportion of heterozygous loci. Maximizes the expected proportion of heterozygous loci in offspring produced from random crossings within the selected core. In contrast to Shannon's index (SH) this objective treats each marker (locus) with equal importance, regardless of the number of possible alleles for that marker. Requires genotypes.

CV

Allele coverage. Maximizes the proportion of alleles observed in the full dataset that are retained in the selected core. Requires genotypes.

The first three objective types (EN, AN and EE) aggregate pairwise distances between individuals. These distances can be computed using various measures:

MR

Modified Rogers distance (default). Requires genotypes.

CE

Cavalli-Sforza and Edwards distance. Requires genotypes.

GD

Gower distance. Requires phenotypes.

PD

Precomputed distances. Uses the precomputed distance matrix of the dataset.

Usage

objective(
  type = c("EN", "AN", "EE", "SH", "HE", "CV"),
  measure = c("MR", "CE", "GD", "PD"),
  weight = 1,
  range = NULL
)

Arguments

type

Objective type, one of EN (default), AN, EE, SH, HE or CV (see description). The former three objectives are distance based and require to choose a distance measure. By default, Modified Roger's distance is used, computed from the genotypes.

measure

Distance measure used to compute the distance between two individuals, one of MR (default), CE, GD or PD (see description). Ignored when type is SH, HE or CV.

weight

Weight assigned to the objective when maximizing a weighted index. Defaults to 1.0.

range

Normalization range [l,u] of the objective when maximizing a weighted index. By default the range is not set (NULL) and will be determined automatically prior to execution, if normalization is enabled (default). Values are rescaled to [0,1] with the linear formula v' = (v - l)/(u - l) . When an explicit normalization range is set, it overrides the automatically inferred range. Also, setting the range for all included objectives reduces the computation time when sampling a multi-objective core collection. In case of repeated sampling from the same dataset with the same objectives and size, it is therefore advised to determine the normalization ranges only once using getNormalizationRanges so that they can be reused for all executions.

Value

Core Hunter objective of class chobj with elements

type

Objective type.

meas

Distance measure (if applicable).

weight

Assigned weight.

range

Normalization range (if specified).

See Also

getNormalizationRanges, setRange

Examples

objective()
objective(meas = "PD")
objective("EE", "GD")
objective("HE")
objective("EN", "MR", range = c(0.150, 0.300))
objective("AN", "MR", weight = 0.5, range = c(0.150, 0.300))