Convex Combination Recombinator

Numeric Values between various individuals are recombined via component-wise convex combination (or weighted mean). The number of individuals over which the convex combination is taken must be determined during construction as n_indivs_in.

The number of output individuals is always 1, i.e. n_indivs_in are used to create one output value. When using this recombinator in a typical EA setting, e.g. with mies_generate_offspring, it is therefore recommended to use a parent-selector where the expected quality of selected parents does not depend on the number of parents selected when n_indivs_in is large: sel("tournament") is preferred to sel("best").

Configuration Parameters

• lambda :: numeric | matrix
  Combination weights; these are normalized to sum to 1 internally. Must either be a vector of length n_indivs_in, or a matrix with n_indivs_in rows and as many columns as there are components in the values being operated on. Must be non-negative, at least one value per column must be greater than zero, but it is not necessary that they sum to 1.
  Initialized to rep(1, n_indivs_in), i.e. equal weights to all individuals being operated on.

Supported Operand Types

Supported Param classes are: ParamDbl

Dictionary

This Recombinator can be created with the short access form rec() (recs() to get a list), or through the the dictionary dict_recombinators in the following way:

# preferred:
rec("convex")
recs("convex")  # takes vector IDs, returns list of Recombinators

# long form:
dict_recombinators$get("convex")

See also

Other recombinators: OperatorCombination, RecombinatorPair, Recombinator, dict_recombinators_cmpmaybe, dict_recombinators_cvxpair, dict_recombinators_maybe, dict_recombinators_null, dict_recombinators_proxy, dict_recombinators_sbx, dict_recombinators_sequential, dict_recombinators_swap, dict_recombinators_xonary, dict_recombinators_xounif

Super classes

miesmuschel::MiesOperator -> miesmuschel::Recombinator -> RecombinatorConvex

Methods

Public methods

• RecombinatorConvex$new()

• RecombinatorConvex$clone()

Inherited methods

• miesmuschel::MiesOperator$help()
• miesmuschel::MiesOperator$operate()
• miesmuschel::MiesOperator$prime()
• miesmuschel::MiesOperator$print()
• miesmuschel::MiesOperator$repr()

---

Method new()

Initialize the RecombinatorConvex object.

Usage

RecombinatorConvex$new(n_indivs_in = 2)

Arguments

n_indivs_in

(integer(1))
Number of individuals to consider at the same time. When operating, the number of input individuals must be divisible by this number. Default 2.
The $n_indivs_in field will reflect this value.

---

Method clone()

The objects of this class are cloneable with this method.

Usage

RecombinatorConvex$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

set.seed(1)
rcvx = rec("convex", n_indivs_in = 3)
p = ps(x = p_dbl(-5, 5), y = p_dbl(-5, 5), z = p_dbl(-5, 5))
data = data.frame(x = 0:5, y = 0:5, z = 0:5)

rcvx$prime(p)
rcvx$operate(data)  # mean of groups of 3
#>   x y z
#> 1 1 1 1
#> 2 4 4 4

rcvx = rec("convex", 3, lambda = c(0, 1, 2))$prime(p)
rcvx$operate(data)  # for groups of 3, take 1/3 of 2nd and 2/3 of 3rd row
#>          x        y        z
#> 1 1.666667 1.666667 1.666667
#> 2 4.666667 4.666667 4.666667

lambda = matrix(c(0, 1, 2, 1, 1, 1, 1, 0, 0), ncol = 3)
lambda
#>      [,1] [,2] [,3]
#> [1,]    0    1    1
#> [2,]    1    1    0
#> [3,]    2    1    0

rcvx = rec("convex", 3, lambda = lambda)$prime(p)
rcvx$operate(data)  # componentwise different operation
#>          x y z
#> 1 1.666667 1 0
#> 2 4.666667 4 3