| Title: | Optimisation with Particle Swarm Optimisation |
|---|---|
| Description: | A toolbox to create a particle swarm optimisation (PSO), the package contains two classes: the Particle and the Particle Swarm, this two class is used to run the PSO with methods to easily print, plot and save the result. |
| Authors: | Theo Brunet [aut, cre] |
| Maintainer: | Theo Brunet <[email protected]> |
| License: | GPL-3 |
| Version: | 1.0 |
| Built: | 2025-02-21 04:00:08 UTC |
| Source: | https://github.com/cran/particle.swarm.optimisation |
Class for the Particles used in the Particle Swarm Optimisation, It is call by the Particle Swarm object to make the population.
values_ranges(list) max and min for each value of the particle
values(numeric) values of the particle (his position in space)
fitness(numeric) fitness of the particle (his score)
fitness_function(function) function used to find the fitness
personal_best_values(numeric) Best values of the particle
personal_best_fitness(numeric) Fitness of the best values
velocity(numeric) Velocity of the particle (one velocity for each values)
acceleration_coefficient(numeric) coefficient c1 and c2 (for personal and global best)
inertia(numeric) inertia of the particle
new()
Create a new Particle object.
Particle$new( values_ranges, values, fitness_function, acceleration_coefficient, inertia )
values_rangesrange for each value of the particle (min and max), his size need to be the same as values. (List)
values, values of the particles. (numeric)
fitness_functionfunction used to test the Particle and find his fitness. (function)
acceleration_coefficienta vector of two values, one for c1 (the personal coefficient), and one for c2 (the global coefficient). (numeric)
inertiaThe inertia of the particle (the influence of the previous velocity on the next velocity). (numeric)
A new Particle object.
get_fitness()
Calculate the fitness of the particle with the fitness function and save it in self$fitness
Particle$get_fitness()
self
update()
Update Particle's position and velocity.
Particle$update(swarm_best)
swarm_bestthe best values of the swarm used to update the velocity
self
update_personal_best_fitness()
Update the Particle's best values and fitness.
Particle$update_personal_best_fitness()
self
print()
print the current values of the particle and his fitness
Particle$print()
clone()
The objects of this class are cloneable with this method.
Particle$clone(deep = FALSE)
deepWhether to make a deep clone.
# If you use the Particle Swarm Object there is no need to manually create the Particle # But if you want to use the Particle for another project: # In this example we use the PSO to solve the following equation: # a * 5 + b * 25 + 10 = 15 fitness_function <- function(values){ a <- values[1] b <- values[2] particule_result <- a*5 + b*25 + 10 difference <- 15 - particule_result fitness <- 1 - abs(difference) return(fitness) } values_ranges <- list(c(-10^3,10^3),c(-10^3,10^3)) particle_example <- Particle$new(values_ranges = values_ranges, values = c(0,0), fitness_function = fitness_function, acceleration_coefficient = c(0.5,0.5), inertia = 0.4) print(particle_example) particle_example$get_fitness() print(particle_example) particle_example$update(c(10,25)) print(particle_example)# If you use the Particle Swarm Object there is no need to manually create the Particle # But if you want to use the Particle for another project: # In this example we use the PSO to solve the following equation: # a * 5 + b * 25 + 10 = 15 fitness_function <- function(values){ a <- values[1] b <- values[2] particule_result <- a*5 + b*25 + 10 difference <- 15 - particule_result fitness <- 1 - abs(difference) return(fitness) } values_ranges <- list(c(-10^3,10^3),c(-10^3,10^3)) particle_example <- Particle$new(values_ranges = values_ranges, values = c(0,0), fitness_function = fitness_function, acceleration_coefficient = c(0.5,0.5), inertia = 0.4) print(particle_example) particle_example$get_fitness() print(particle_example) particle_example$update(c(10,25)) print(particle_example)
Particle Swarm, used to launch the Particle Swarm Optimisation, The PSO is used to maximise the fitness.
pop_size(numeric) number of particles in the swarm
ranges_of_values(list) range for each value for the particle
values_names(list) list of names for each value (optionnal)
pop(list) list of particle in the swarm
fitness_function(function) fitness function used to find the fitness of the particle
list_fitness(list) list of fitness of the particles
max_it(numeric) maximum number of iteration
acceleration_coefficient_range(list) coefficient c1 and c2 for the particles
swarm_best_fitness(numeric) best fitness of the swarm
swarm_best_values(numeric) values of the particle with the best fitness
inertia(numeric) inertia of the particles
new()
Create a new ParticleSwarm object.
ParticleSwarm$new( pop_size, values_names, fitness_function, max_it, acceleration_coefficient_range, inertia, ranges_of_values )
pop_sizenumber of individu in the swarm. (numeric)
values_nameslist of names for each value (character)
fitness_functionfunction used to test the Particle and find his fitness. (function)
max_itMaximum number of iteration for the PSO. (numeric)
acceleration_coefficient_rangea vector of four values (min and max for c1 and c2) (numeric)
inertiaThe inertia for the particle (the influence of the previous velocity on the next velocity). (numeric)
ranges_of_valuesrange for each value of the particle (min and max). (List)
A new ParticleSwarm object.
# Create a ParticleSwarm object
swarm <- ParticleSwarm$new(pop_size=20,
values_names=c('a','b'),
max_it=20,
fitness_function = function(values){return(values[1]+values[2])},
acceleration_coefficient=list(c(0.5,1),c(0.5,1)),
inertia=0.5,
ranges_of_values=list(c(-100,100),c(-100,100)))
run()
Make the Particle Swarm Optimisation
ParticleSwarm$run( verbose = TRUE, plot = TRUE, save_file = FALSE, dir_name = "PSO_pop" )
verboseprint the different step (iteration and individu)
plotplot the result of each iteration (only for 2D or 3D problem)
save_filesave the population of each Iteration in a file and save the plot if plot=TRUE
dir_namename of the directory, default value is PSO_pop
self
# Create a ParticleSwarm object
swarm <- ParticleSwarm$new(pop_size=20,
values_names=c('a','b'),
max_it=20,
fitness_function = function(values){return(values[1]+values[2])},
acceleration_coefficient=list(c(0.5,1),c(0.5,1)),
inertia=0.5,
ranges_of_values=list(c(-100,100),c(-100,100)))
# run the PSO
swarm$run(verbose = FALSE,
plot = FALSE,
save_file = FALSE)
# return the best result:
print(swarm$swarm_best_values)
generate_pop()
create the population of the swarm (this method is automatically called by the run method)
ParticleSwarm$generate_pop(verbose = TRUE)
verboseprint the advancement or not
self
move_the_swarm()
The method used to change the location of each particle (this method is automatically called by the run method)
ParticleSwarm$move_the_swarm(verbose)
verboseprint or not the advancement
self
save_pop()
The method used to save the values and fitness of the population in a CSV file (this method is automatically called by the run method if you have chosen to save the result)
ParticleSwarm$save_pop(nb_it, dir_name)
nb_itnumber of the iteration, used to create the name of the csv file
dir_nameName of the directory
self
plot_the_swarm_2D()
method used to plot a 2D plot (this method is automatically called by the run method if you have chosen to plot the swarm)
ParticleSwarm$plot_the_swarm_2D(nb_it, save_file)
nb_itnumber of the iteration used to save the plot as a png
save_filesave the plot as a file
self
plot_the_swarm_3D()
method used to plot a 3D plot
ParticleSwarm$plot_the_swarm_3D(nb_it, save_file)
nb_itnumber of the iteration used to save the plot as a png (this method is automatically called by the run method if you have chosen to plot the swarm)
save_filesave the plot as a file
self
print()
Print the current result of the population
ParticleSwarm$print()
clone()
The objects of this class are cloneable with this method.
ParticleSwarm$clone(deep = FALSE)
deepWhether to make a deep clone.
# In this example we use the PSO to solve the following equation: # a * 5 + b * 25 + 10 = 15 fitness_function <- function(values){ a <- values[1] b <- values[2] particule_result <- a*5 + b*25 + 10 difference <- 15 - particule_result fitness <- 1 - abs(difference) return(fitness) } values_ranges <- list(c(-10^3,10^3),c(-10^3,10^3)) swarm <- ParticleSwarm$new(pop_size = 200, values_names = list("a","b"), fitness_function = fitness_function, max_it = 75, acceleration_coefficient_range = list(c(0,1),c(0,1)), inertia = 0.5, ranges_of_values = values_ranges) swarm$run(plot = FALSE,verbose = FALSE,save_file = FALSE) # the solution is : swarm$swarm_best_values swarm$swarm_best_values[[1]]*5 + swarm$swarm_best_values[[2]] *25 + 10 ## ------------------------------------------------ ## Method `ParticleSwarm$new` ## ------------------------------------------------ # Create a ParticleSwarm object swarm <- ParticleSwarm$new(pop_size=20, values_names=c('a','b'), max_it=20, fitness_function = function(values){return(values[1]+values[2])}, acceleration_coefficient=list(c(0.5,1),c(0.5,1)), inertia=0.5, ranges_of_values=list(c(-100,100),c(-100,100))) ## ------------------------------------------------ ## Method `ParticleSwarm$run` ## ------------------------------------------------ # Create a ParticleSwarm object swarm <- ParticleSwarm$new(pop_size=20, values_names=c('a','b'), max_it=20, fitness_function = function(values){return(values[1]+values[2])}, acceleration_coefficient=list(c(0.5,1),c(0.5,1)), inertia=0.5, ranges_of_values=list(c(-100,100),c(-100,100))) # run the PSO swarm$run(verbose = FALSE, plot = FALSE, save_file = FALSE) # return the best result: print(swarm$swarm_best_values)# In this example we use the PSO to solve the following equation: # a * 5 + b * 25 + 10 = 15 fitness_function <- function(values){ a <- values[1] b <- values[2] particule_result <- a*5 + b*25 + 10 difference <- 15 - particule_result fitness <- 1 - abs(difference) return(fitness) } values_ranges <- list(c(-10^3,10^3),c(-10^3,10^3)) swarm <- ParticleSwarm$new(pop_size = 200, values_names = list("a","b"), fitness_function = fitness_function, max_it = 75, acceleration_coefficient_range = list(c(0,1),c(0,1)), inertia = 0.5, ranges_of_values = values_ranges) swarm$run(plot = FALSE,verbose = FALSE,save_file = FALSE) # the solution is : swarm$swarm_best_values swarm$swarm_best_values[[1]]*5 + swarm$swarm_best_values[[2]] *25 + 10 ## ------------------------------------------------ ## Method `ParticleSwarm$new` ## ------------------------------------------------ # Create a ParticleSwarm object swarm <- ParticleSwarm$new(pop_size=20, values_names=c('a','b'), max_it=20, fitness_function = function(values){return(values[1]+values[2])}, acceleration_coefficient=list(c(0.5,1),c(0.5,1)), inertia=0.5, ranges_of_values=list(c(-100,100),c(-100,100))) ## ------------------------------------------------ ## Method `ParticleSwarm$run` ## ------------------------------------------------ # Create a ParticleSwarm object swarm <- ParticleSwarm$new(pop_size=20, values_names=c('a','b'), max_it=20, fitness_function = function(values){return(values[1]+values[2])}, acceleration_coefficient=list(c(0.5,1),c(0.5,1)), inertia=0.5, ranges_of_values=list(c(-100,100),c(-100,100))) # run the PSO swarm$run(verbose = FALSE, plot = FALSE, save_file = FALSE) # return the best result: print(swarm$swarm_best_values)