Certara.RDarwin (version 1.2.0)

pyDarwinOptionsMOGA: Create Options for the pyDarwin MOGA Block

Description

Generates a list of specific options for the MOGA (Multi-Objective Genetic Algorithm) or MOGA3 (NSGA-III) algorithms in pyDarwin. This list is intended to be passed as the MOGA argument to the create_pyDarwinOptions() function when algorithm is set to "MOGA" or "MOGA3".

Usage

pyDarwinOptionsMOGA(
  objectives = 3,
  names = NULL,
  constraints = 0,
  partitions = 12,
  crossover = "single",
  crossover_rate = 0.95,
  mutation_rate = 0.95,
  attribute_mutation_probability = 0.1
)

Value

A list containing MOGA-specific options.

Arguments

objectives

Positive integer: Number of objectives. Applicable only when the algorithm in create_pyDarwinOptions() is set to "MOGA3". If the algorithm is "MOGA" (implying NSGA-II), this parameter is ignored by pyDarwin, and 2 objectives (OFV and NEP) are used internally. For "MOGA3", objectives are defined by postprocessing. Default: 3 (relevant for "MOGA3").

names

Character vector (optional): List of names for the objectives. Applicable only when the algorithm is "MOGA3". The length of this vector should match the objectives value. If NULL, empty, or of a different size, pyDarwin uses generic names (e.g., "f1", "f2", "f3"). These names are used for reporting in results.csv. Ignored if the algorithm is "MOGA". Default: NULL.

constraints

Non-negative integer: Number of constraints. Applicable only when the algorithm is "MOGA3". Constraints must be provided by user-defined postprocessing scripts (R or Python). See https://pymoo.org/constraints/index.html. Ignored if the algorithm is "MOGA". Default: 0.

partitions

Positive integer: Number of partitions for the reference directions used in NSGA-III. Applicable only when the algorithm is "MOGA3". See https://pymoo.org/misc/reference_directions.html. Ignored if the algorithm is "MOGA". Default: 12.

crossover

Character string: Crossover algorithm. When set to "single", SinglePointCrossover is used. Otherwise, for other values like "two_point", TwoPointCrossover is typically used by pymoo. See https://pymoo.org/operators/crossover.html#Point-Crossover. Applicable for both "MOGA" and "MOGA3". Default: "single".

crossover_rate

Numeric value between 0.0 and 1.0: The fraction of mating pairs that will undergo crossover. Applicable for both "MOGA" and "MOGA3". Default: 0.95.

mutation_rate

Numeric value between 0.0 and 1.0: The probability that at least one bit in the genome will be "flipped" (mutated). Applicable for both "MOGA" and "MOGA3". Default: 0.95.

attribute_mutation_probability

Numeric value between 0.0 and 1.0: The probability of any individual bit (attribute) in the genome being mutated. Applicable for both "MOGA" and "MOGA3". Default: 0.1.

Details

This function defines parameters for the MOGA settings block in pyDarwin. The relevance of certain parameters (objectives, names, constraints, partitions) depends on whether the algorithm in create_pyDarwinOptions() is set to "MOGA" (for NSGA-II) or "MOGA3" (for NSGA-III).

• If algorithm = "MOGA" (NSGA-II): pyDarwin internally uses 2 objectives (OFV and NEP). The objectives, names, constraints, and partitions parameters from this MOGA options block are ignored by pyDarwin. Postprocessing scripts are not used by pyDarwin to define objectives.

• If algorithm = "MOGA3" (NSGA-III): The objectives, names, constraints, and partitions parameters from this MOGA options block are utilized by pyDarwin. User-supplied postprocessing scripts (R or Python) are required to calculate and return the values for the objectives and constraints. (R: list of two vectors; Python: tuple of two lists).

Common parameters like crossover, crossover_rate, mutation_rate, and attribute_mutation_probability apply to both "MOGA" and "MOGA3" variants.

Examples

# MOGA options, defaults are generally suitable for algorithm = "MOGA3"
# if postprocessing handles 3 objectives.
moga_block_for_moga3 <- pyDarwinOptionsMOGA(
  objectives = 3,
  names = c("Objective1", "Objective2", "Objective3"),
  constraints = 1,
  partitions = 10
)

# MOGA options where specific settings are for NSGA-II (algorithm = "MOGA")
# Note: objectives, names, constraints, partitions would be ignored by pyDarwin.
moga_block_for_moga <- pyDarwinOptionsMOGA(
  crossover = "two_point",
  crossover_rate = 0.92
)