Package 'hmsr'

Title: Multipopulation Evolutionary Strategy HMS
Description: The HMS (Hierarchic Memetic Strategy) is a composite global optimization strategy consisting of a multi-population evolutionary strategy and some auxiliary methods. The HMS makes use of a dynamically-evolving data structure that provides an organization among the component populations. It is a tree with a fixed maximal height and variable internal node degree. Each component population is governed by a particular evolutionary engine. This package provides a simple R implementation with examples of using different genetic algorithms as the population engines. References: J. Sawicki, M. Łoś, M. Smołka, J. Alvarez-Aramberri (2022) <doi:10.1007/s11047-020-09836-w>.
Authors: Wojciech Achtelik [aut, cre], Marcin Kozubek [aut], Maciej Smołka [ths, aut] (<https://orcid.org/0000-0002-3386-0555>, Java original), AGH University of Kraków [cph]
Maintainer: Wojciech Achtelik <[email protected]>
License: MIT + file LICENSE
Version: 1.0.1
Built: 2025-02-06 04:42:46 UTC
Source: https://github.com/wojtacht/hms

Help Index

Calculate selected ELA features for given fitness.

Description

Calculate selected ELA features for given fitness.

Usage

calculate_ela_features(fitness, lower, upper)

Arguments

fitness

fitness function, that returns a numerical value.
lower

numeric - lower bound of the domain, a vector of length equal to the decision variables.
upper

numeric - upper bound of the domain, a vector of length equal to the decision variables.

Value

Returns a named list with values of selected ELA features.

Classify optimization problem using selected ELA features and Random Forest model trained on BBOB dataset.

Description

Classify optimization problem using selected ELA features and Random Forest model trained on BBOB dataset.

Usage

classify_optimization_problem(fitness, lower, upper)

Arguments

fitness

fitness function, that returns a numerical value, to be classified. The domain should be at least two dimensional.
lower

numeric - lower bound of the domain, a vector of length equal to the decision variables.
upper

numeric - upper bound of the domain, a vector of length equal to the decision variables.

Value

Returns one of c("low-conditioning", "multimodal-adequate", "multimodal-weak", "separable", "unimodal").

Examples

f <- function(x) x[[1]] + x[[2]]
result <- classify_optimization_problem(fitness = f, lower = c(-5, -5), upper = c(5, 5))

Function that runs one cmaes metaepoch. Wrapper function for cmaes::cma_es.

Description

Function that runs one cmaes metaepoch. Wrapper function for cmaes::cma_es.

Usage

cma_es_metaepoch(config_cmaes)

Arguments

config_cmaes

  • list of cmaes::cma_es params

Value

list with named fields: solution, population, value. See ga_metaepoch for more details.

Examples

tree_height <- 3
empty_config_cma_es <- lapply(1:tree_height, function(x) {
  list()
})
cma_es_metaepoch(empty_config_cma_es)

Function that runs gradient method for one deme. Wrapper function for stats::optim.

Description

Function that runs gradient method for one deme. Wrapper function for stats::optim.

Usage

default_run_gradient_method(deme, fitness, optim_args)

Arguments

deme

  • Deme

fitness

  • fitness function

optim_args

  • list of additional parameters (stats::optim parameters)

Value

list with named fields: solution, population, value. See ga_metaepoch for more details.

Function that generates run_metaepoch function for two level HMS. First level: DE, second level: CMA-ES.

Description

Function that generates run_metaepoch function for two level HMS. First level: DE, second level: CMA-ES.

Usage

deoptim_cma_es_metaepoch(deoptim_cma_es_config)

Arguments

deoptim_cma_es_config

  • list that consists of two lists: DEoptim params and CMA-ES params.

Value

list with named fields: solution, population, value. See ga_metaepoch for more details.

Examples

tree_height <- 2
de_config <- list()
cma_es_config <- list()
config <- list(de_config, cma_es_config)
deoptim_cma_es_metaepoch(config)

Function that runs one differential evolution metaepoch. Wrapper function for DEoptim::DEoptim.

Description

Function that runs one differential evolution metaepoch. Wrapper function for DEoptim::DEoptim.

Usage

deoptim_metaepoch(config_deoptim)

Arguments

config_deoptim

  • list of DEoptim::DEoptim params

Value

list with named fields: solution, population, value. See ga_metaepoch for more details.

Examples

tree_height <- 3
empty_config_deoptim <- lapply(1:tree_height, function(x) {
  list()
})
deoptim_metaepoch(empty_config_deoptim)

Function that runs one ecr metaepoch. Wrapper function for ecr::ecr.

Description

Function that runs one ecr metaepoch. Wrapper function for ecr::ecr.

Usage

ecr_metaepoch(config_ecr)

Arguments

config_ecr

  • list of ecr::ecr params

Value

list with named fields: solution, population, value, fitness_values. See ga_metaepoch for more details.

Examples

tree_height <- 3
empty_config_ecr <- lapply(1:tree_height, function(x) {
  list()
})
ecr_metaepoch(empty_config_ecr)

Euclidean distance

Description

Euclidean distance

Usage

euclidean_distance(x, y)

Arguments

x

  • numeric

y

  • numeric

Value

numeric - euclidean distance between x and y

Examples

euclidean_distance(c(1, 1), c(1, 2))

Function that generates run_metaepoch function for two level HMS. First level: GA, second level: CMA-ES.

Description

Function that generates run_metaepoch function for two level HMS. First level: GA, second level: CMA-ES.

Usage

ga_cma_es_metaepoch(ga_cma_es_config)

Arguments

ga_cma_es_config

  • list that consists of two lists: GA params and CMA-ES params.

Value

list with named fields: solution, population, value. See ga_metaepoch for more details.

Examples

tree_height <- 2
ga_config <- list()
cma_es_config <- list()
config <- list(ga_config, cma_es_config)
ga_cma_es_metaepoch(config)

Function that runs one GA metaepoch. Wrapper function for GA::ga.

Description

Function that runs one GA metaepoch. Wrapper function for GA::ga.

Usage

ga_metaepoch(config_ga)

Arguments

config_ga

  • list of GA::ga params

Value

list with named fields: solution, population, value, fitness_values, context or NULL. A solution is a value of the decision variable giving the best fitness. A population is a matrix representing final population. A value is the value of a fitness function for the solution. A fitness_values is a vector of fitness values for the final population. A context is a list with internal state of the metaepoch (or NULL if it's not necessary). NULL can be returned if GA::ga fails and "ignore_errors" is TRUE in the config.

Examples

tree_height <- 3
empty_config_ga <- lapply(1:tree_height, function(x) {
  list("ignore_errors" = TRUE)
})
ga_metaepoch(empty_config_ga)

Factory function for a global stopping condition that stops the computation after fitness function has been evaluated given number of times.

Description

Factory function for a global stopping condition that stops the computation after fitness function has been evaluated given number of times.

Usage

gsc_max_fitness_evaluations(max_evaluations)

Arguments

max_evaluations

  • numeric - maximum number of fitness function evaluations

Value

Function that receives a list of metaepoch snapshots and returns a Boolean value determining whether the computation should be stopped based on how many fitness function evaluations have been made, which can be used as a global stopping condition for the hms function.

Examples

global_stopping_condition <- gsc_max_fitness_evaluations(10000)

Factory function for a global stopping condition that stops the computation after given number of metaepochs.

Description

Factory function for a global stopping condition that stops the computation after given number of metaepochs.

Usage

gsc_metaepochs_count(metaepochs_count)

Arguments

metaepochs_count

  • numeric - maximum number of metaepochs

Value

Function that receives a list of metaepoch snapshots and returns a Boolean value determining whether the computation should be stopped based on how many metaepochs have passed, which can be used as a global stopping condition for the hms function.

Examples

global_stopping_condition <- gsc_metaepochs_count(10)

Factory function for a global stopping condition that never stops the computation. It results in hms running until there are no more active demes.

Description

Factory function for a global stopping condition that never stops the computation. It results in hms running until there are no more active demes.

Usage

gsc_trivial()

Value

function that always returns FALSE, which can be used as a global stopping condition for the hms function.

Examples

global_stopping_condition <- gsc_trivial()

Maximization (or minimization) of a fitness function using Hierarchic Memetic Strategy.

Description

Maximization (or minimization) of a fitness function using Hierarchic Memetic Strategy.

Usage

hms(
  tree_height = 3,
  minimize = FALSE,
  fitness,
  lower,
  upper,
  sigma = default_sigma(lower, upper, tree_height),
  population_sizes = default_population_sizes(tree_height),
  run_metaepoch = default_ga_metaepoch(tree_height),
  gsc = gsc_default,
  lsc = lsc_default,
  sc = sc_max_metric(euclidean_distance, sprouting_default_euclidean_distances(sigma)),
  create_population = default_create_population(sigma),
  suggestions = NULL,
  with_gradient_method = FALSE,
  gradient_method_args = default_gradient_method_args,
  run_gradient_method,
  monitor_level = "basic",
  parallel = FALSE,
  use_memoise = FALSE
)

Arguments

tree_height

numeric - default value: 5. It determines the maximum tree height which will usually be reached unless a very strict local stopping condition, global stopping condition or sprouting condition is used.
minimize

logical - TRUE when fitness shall be minimized.
fitness

fitness function, that returns a numerical value, to be optimized by the strategy.
lower

numeric - lower bound of the domain, a vector of length equal to the decision variables.
upper

numeric - upper bound of the domain, a vector of length equal to the decision variables.
sigma

numeric - Vector of standard deviations for each tree level used to create a population of a sprouted deme.
population_sizes

numeric - Sizes of deme populations on each tree level.
run_metaepoch

A function that takes 5 parameters: fitness, suggestions, lower, upper, tree_level, runs a metaepoch on the given deme population and returns list with 3 named fields: solution, population, value.
gsc

global stopping condition function taking a list of MetaepochSnapshot objects and returning a logical value; it is evaluated after every metaepoch and determines whether whole computation should be stopped. See gsc_metaepochs_count for more details.
lsc

local stopping condition - function taking a deme and a list of MetaepochSmapshot objects representing previous metaepochs; it is run on every deme after it has run a metaepoch and determines whether that deme will remain active. See lsc_max_fitness_evaluations for more details.
sc

sprouting condition - function taking 3 arguments: an individual, a tree level and a list of Deme objects; it determines whether the given individual can sprout a new deme on the given level. See sc_max_metric for more details.
create_population

function taking 6 parameters: mean, lower, upper, population_size, tree_level, sigma that returns a population for a Deme object to be created on the given tree level.
suggestions

matrix of individuals for the initial population of the root
with_gradient_method

logical determining whether a gradient method should be run for all leaves at the end of the computation to refine their best solutions.
gradient_method_args

list of parameters that are passed to the gradient method
run_gradient_method

function - returns list with named fields: solution, population, value
monitor_level

string - one of: 'none', 'basic', 'basic_tree', 'verbose_tree'.
parallel

logical - TRUE when run_metaepoch runs in parallel.
use_memoise

logical - FALSE when memoise package shall be used to cache fitness function values.

Value

Returns an object of class hms.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)

A S4 class representing a result of hms.

Description

A S4 class representing a result of hms.

Slots

root_id

character - UUID of a root Deme.

metaepoch_snapshots

list of objects of class MetaepochSnapshot.

best_fitness

numeric - best fitness value of all metaepochs.

best_solution

numeric - best solution of all metaepochs.

total_time_in_seconds

numeric - time of a hms execution in seconds.

total_metaepoch_time_in_seconds

numeric - time of all metaepochs in seconds.

metaepochs_count

numeric - total number of all metaepochs.

deme_population_sizes

numeric - sizes of deme populations on each tree level. Same as population_sizes parameter of hms function.

lower

numeric - lower bound of the domain, a vector of length equal to the decision variables.

upper

numeric - upper bound of the domain, a vector of length equal to the decision variables.

call

language - an object of class "call" representing the matched call.

Factory function for a local stopping condition that stops a deme after given number of fitness function evaluations has been made in that deme.

Description

Factory function for a local stopping condition that stops a deme after given number of fitness function evaluations has been made in that deme.

Usage

lsc_max_fitness_evaluations(max_evaluations)

Arguments

max_evaluations

  • numeric

Value

Function that can be used as a local stopping condition for hms.

Examples

local_stopping_condition <- lsc_max_fitness_evaluations(500)

Factory function for a local stopping condition that stops a deme after given number of metaepochs have past since last metaepoch during which this deme had an active child.

Description

Factory function for a local stopping condition that stops a deme after given number of metaepochs have past since last metaepoch during which this deme had an active child.

Usage

lsc_metaepochs_without_active_child(metaepochs_limit)

Arguments

metaepochs_limit

  • number of metaepochs that a deme can be active without any active child

Value

Function that can be used as a local stopping condition for hms.

Examples

local_stopping_condition <- lsc_metaepochs_without_active_child(3)

Factory function for a local stopping condition that stops a deme after given number of consecutive metaeopochs without an improvement of the best solution found in that deme.

Description

Factory function for a local stopping condition that stops a deme after given number of consecutive metaeopochs without an improvement of the best solution found in that deme.

Usage

lsc_metaepochs_without_improvement(max_metaepochs_without_improvement)

Arguments

max_metaepochs_without_improvement

  • numeric

Value

Function that can be used as a local stopping condition for hms.

Examples

local_stopping_condition <- lsc_metaepochs_without_improvement(5)

Factory function for a trivial local stopping condition that lets a deme be active forever. It is usually used in the root of a hms tree.

Description

Factory function for a trivial local stopping condition that lets a deme be active forever. It is usually used in the root of a hms tree.

Usage

lsc_trivial()

Value

Function that always returns FALSE, which can be used as a local stopping condition for hms.

Examples

local_stopping_condition <- lsc_trivial()

Manhattan distance

Description

Manhattan distance

Usage

manhattan_distance(x, y)

Arguments

x

  • numeric

y

  • numeric

Value

numeric - manhattan distance between x and y

Examples

manhattan_distance(c(1, 1), c(1, 2))

A S4 class representing a snapshot of one metaepoch.

Description

A S4 class representing a snapshot of one metaepoch.

Slots

demes

list of objects of class Deme.

best_fitness

numeric - best fitness value of a metaepoch.

best_solution

numeric - best solution of a metaepoch.

time_in_seconds

numeric - time of metaepoch in seconds.

fitness_evaluations

numeric - number of fitness evaluations.

blocked_sprouts

list - list of sprouts that were blocked by sprouting condition. A sprout is a potential origin of a new Deme, it can be blocked by sc – sprouting condition. See sc_max_metric for more details.

is_evolutionary

logical - TRUE for all metaepochs except the gradient one.

Plot method for "hms" class.

Description

Plot method for "hms" class.

Usage

## S4 method for signature 'hms'
plot(x)

Arguments

x

  • hms s4 object

Value

It doesn't return anything meaningful. It plots the fitness by metaepoch count.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
plot(result)

plotActiveDemes method for "hms" class.

Description

plotActiveDemes method for "hms" class.

Usage

plotActiveDemes(object)

Arguments

object

  • hms s4 object

Value

It doesn't return anything meaningful. It plots the number of active demes per metaepoch.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
plotActiveDemes(result)

plotActiveDemes method for "hms" class.

Description

plotActiveDemes method for "hms" class.

Usage

## S4 method for signature 'hms'
plotActiveDemes(object)

Arguments

object

  • hms s4 object

Value

It doesn't return anything meaningful. It plots the number of active demes per metaepoch.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
plotActiveDemes(result)

plotPopulation method for "hms" class.

Description

plotPopulation method for "hms" class.

Usage

plotPopulation(object, dimensions)

Arguments

object

  • hms s4 object

dimensions

  • two selected dimensions

Value

It doesn't return anything meaningful. It plots the selected two dimensions of a population.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
plotPopulation(result, c(1, 1))

plotPopulation method for "hms" class.

Description

plotPopulation method for "hms" class.

Usage

## S4 method for signature 'hms'
plotPopulation(object, dimensions)

Arguments

object

  • hms s4 object

dimensions

  • two selected dimensions

Value

It doesn't return anything meaningful. It plots the selected two dimensions of a population.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
plotPopulation(result, c(1, 1))

Print method for class "hms".

Description

Print method for class "hms".

Usage

## S4 method for signature 'hms'
print(x, ...)

Arguments

x

  • hms s4 object

...

  • other print arguments

Value

It does not return anything. The obvious side effect is output to the terminal.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
print(result)

printBlockedSprouts method for "hms" class.

Description

printBlockedSprouts method for "hms" class.

Usage

printBlockedSprouts(object)

Arguments

object

  • hms s4 object

Value

It doesn't return anything. It prints blocked sprouts per metaepoch.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
printBlockedSprouts(result)

printBlockedSprouts method for "hms" class.

Description

printBlockedSprouts method for "hms" class.

Usage

## S4 method for signature 'hms'
printBlockedSprouts(object)

Arguments

object

  • hms s4 object

Value

It doesn't return anything. It prints blocked sprouts per metaepoch.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
printBlockedSprouts(result)

printTree method for class "hms".

Description

printTree method for class "hms".

Usage

printTree(object)

Arguments

object

  • hms s4 object

Value

It does not return anything. It prints the hms tree.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
printTree(result)

printTree method for class "hms".

Description

printTree method for class "hms".

Usage

## S4 method for signature 'hms'
printTree(object)

Arguments

object

  • hms s4 object

Value

It does not return anything. It prints the hms tree.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
printTree(result)

Factory function that creates normal mutation function

Description

Given the domain bounds and standard deviation returns a function compatible with GA interface that performs a mutation on the given individual using truncated normal distribution.

Usage

rtnorm_mutation(lower, upper, sd)

Arguments

lower

  • Lower bound of the problem's domain

upper

  • Upper bound of the problem's domain

sd

  • Standard deviation of the truncated normal distribution used for the mutation

Value

Function that takes two parameters (the GA object object and an individual to perform the mutation on parent) and returns a new individual that is the result of normal mutation applied to the parent.

Examples

mutation <- rtnorm_mutation(
  lower = rep(-500, 5),
  upper = rep(500, 5),
  sd = rep(50, 5)
)

saveMetaepochsPopulations method for "hms" class.

Description

saveMetaepochsPopulations method for "hms" class.

Usage

saveMetaepochsPopulations(object, path, dimensions)

Arguments

object

hms s4 object
path

path
dimensions

vector of two selected dimensions e.g. c(1,2)

Value

It doesn't return anything. It creates plots and saves them to a specified directory.

Examples

fitness <- function(x) x[1] + x[2]
lower <- c(-5, -5)
upper <- c(5, 5)
result <- hms(fitness = fitness, lower = lower, upper = upper)
selected_dimensions <- c(1, 2)
saveMetaepochsPopulations(result, tempdir(), selected_dimensions)

saveMetaepochsPopulations

Description

saveMetaepochsPopulations

Usage

## S4 method for signature 'hms'
saveMetaepochsPopulations(object, path, dimensions)

Arguments

object

hms s4 object
path

path
dimensions

vector of two selected dimensions e.g. c(1,2)

Value

It doesn't return anything. It creates plots and saves them to a specified directory.

Examples

fitness <- function(x) x[1] + x[2]
lower <- c(-5, -5)
upper <- c(5, 5)
result <- hms(fitness = fitness, lower = lower, upper = upper)
selected_dimensions <- c(1, 2)
saveMetaepochsPopulations(result, tempdir(), selected_dimensions)

Default sprouting condition based on given metric.

Description

It allows an individual to sprout only if there are no other demes on the target level that have centroid within the given distance.

Usage

sc_max_metric(metric, max_distances)

Arguments

metric

  • Metric used for deme distance comparison (e.g. euclidean_distance, manhattan_distance)

max_distances

  • numeric - maximum distance to a centroid of a deme on the target level that would allow the individual to sprout

Value

Function that can be used as a sprouting condition of hms.

Examples

sprouting_condition <- sc_max_metric(euclidean_distance, c(20, 10))

Show method for class "hms".

Description

Show method for class "hms".

Usage

## S4 method for signature 'hms'
show(object)

Arguments

object

  • hms s4 object

Value

It returns the names of the slots and the classes associated with the slots in the "hms" class. It prints call details.

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
show(result)

Summary method for class "hms".

Description

Summary method for class "hms".

Usage

## S4 method for signature 'hms'
summary(object, ...)

Arguments

object

  • hms s4 object

...

  • other summary arguments

Value

Returns a list with fields: fitness, solution, metaepochs, deme_population_sizes, lower_bound, upper_bound, computation_time. These fields should match fields of class "hms".

Examples

f <- function(x) x
result <- hms(fitness = f, lower = -5, upper = 5)
summary(result)

Train Random Forest model on BBOB dataset (data/ela_features.rda).

Description

Train Random Forest model on BBOB dataset (data/ela_features.rda).

Usage

train_random_forest_model()

Value

Returns a Random Forest model trained on ela_features dataset which is stored internally in sysdata.rda. The ela_features dataset, created using the flacco package, includes ELA features for all BBOB functions (all available fids) across different dimensions (2, 5, 10, 20) and instances (1:20). Key features captured are ic.eps.ratio, nbc.nb_fitness.cor, ela_meta.quad_simple.adj_r2, and ela_meta.lin_simple.adj_r2, along with function type.