new GP(params)
Genetic Programming evolutionary search heuristic.
Parameters are:
- Population size -- Size of the population
- Solution size -- Max size of the nodes which represents an individual
solution
- Max depth -- Max depth of a tree, this is a function of the solution size
- Generations -- Number of iterations of the search loop
- Elite size -- Number of individuals preserved between generations
- Crossover probability -- Probability of crossing over two solutions
- Mutation probability -- Probability of mutating a solution
- Fitness function -- Method used to evaluate fitness
- Symbols -- The symbols that a GP tree can be built from
Parameters:
| Name | Type | Description |
|---|---|---|
params |
object | GP parameters |
- Source:
Methods
crossover(crossoverProbability, population) → {Array}
The individuals are created by
selecting two random nodes from two parents and swapping the
subtrees. Checks that the nodes are compatible for crossover, to prevent too
large trees. Loops over the entire population.
Parameters:
| Name | Type | Description |
|---|---|---|
crossoverProbability |
number | Probability to crossover individuals |
population |
array | List of individuals |
- Source:
Returns:
List of crossed over individuals
- Type
- Array
evaluateFitness(population) → {array}
Evaluate the population according to the fitness function.
Parameters:
| Name | Type | Description |
|---|---|---|
population |
array | List of individual solutions |
- Source:
Returns:
Evaluated population
- Type
- array
initializePopulation() → {Array}
Ramped half-half initialization. The individuals in the
population are initialized using the grow or the full method for
each depth value (ramped) up to max_depth.
- Source:
Returns:
Population of individual solutions
- Type
- Array
mutation(mutationProbability, individuals, maxDepth, symbols) → {Array}
Mutate an individual by randomly picking a node and growing a
new subtree from it. Loops over the entire population.
Parameters:
| Name | Type | Description |
|---|---|---|
mutationProbability |
number | Probability to mutate an individual |
individuals |
array | List of individuals |
maxDepth |
number | Max depth of individual |
symbols |
Symbols | Symbols used to label nodes with |
- Source:
Returns:
List of mutated individuals
- Type
- Array
printStats(generation, population) → {Individual}
Print the statistics for the generation and population.
Parameters:
| Name | Type | Description |
|---|---|---|
generation |
number | Generation number |
population |
array | List of individuals |
- Source:
Returns:
Best solution of the generation
- Type
- Individual
searchLoop(population) → {Individual}
Return the best individual from the evolutionary search
loop. Starting from an initial population.
Parameters:
| Name | Type | Description |
|---|---|---|
population |
array | Initial population |
- Source:
Returns:
Best individual solution (model)
- Type
- Individual
tournamentSelection(tournamentSize, population) → {Array}
Return individuals from a population by drawing
`tournamentSize` competitors randomly and selecting the best
of the competitors. `population_size` number of tournaments are
held.
Parameters:
| Name | Type | Description |
|---|---|---|
tournamentSize |
number | Number of competitors in each tournament |
population |
array | List of individuals |
- Source:
Returns:
List of winners from each tournament
- Type
- Array