Petri (petri v0.1.0)

A multi-representation genetic algorithm library.

Petri supports three chromosome encodings:

Petri.Chromosome.Real — continuous values with per-gene bounds
Petri.Chromosome.Permutation — integer permutations (e.g. TSP tours)
Petri.Chromosome.Binary — bit strings for subset selection

Each encoding has its own crossover and mutation operators. Selection, termination, and the generational engine work across all encodings.

Quick start

iex> fitness = fn c -> Petri.Chromosome.genes(c) |> Enum.sum() end
iex> result = Petri.run(fitness, %{
...>   encoding: :binary, length: 10,
...>   population_size: 20, max_generations: 50, seed: 42
...> })
iex> %Petri.Result{} = result
iex> {_chrom, f} = result.best
iex> f
10

Running the examples

Three standalone .exs scripts in the examples/ directory demonstrate each encoding on a realistic problem:

elixir examples/tsp.exs               # Berlin52 TSP (permutation)
elixir examples/ml_hyperparams.exs    # Hyperparameter tuning (real)
elixir examples/feature_selection.exs # Feature subset selection (binary)

Summary

Functions

configure(config)

Validates a config without running the GA.

run(fitness_fn, config)

Runs a genetic algorithm.

Functions

configure(config)

Validates a config without running the GA.

Returns {:ok, config} or {:error, reasons}.

Example

iex> {:ok, config} = Petri.configure(%{
...>   encoding: :binary, length: 8,
...>   population_size: 20, max_generations: 10
...> })
iex> config.encoding
:binary

iex> {:error, err} = Petri.configure(%{
...>   encoding: :binary, crossover: :blx_alpha,
...>   length: 8, population_size: 20, max_generations: 10
...> })
iex> is_list(err)
true

run(fitness_fn, config)

Runs a genetic algorithm.

fitness_fn is a function (chromosome -> fitness) where higher fitness is better. The GA maximizes.

config is a map. Required fields depend on the encoding. See Petri.Config.parse/1 for the full schema.

Config (binary encoding)

iex> fitness = fn c -> Petri.Chromosome.genes(c) |> Enum.sum() end
iex> result = Petri.run(fitness, %{
...>   encoding: :binary, length: 8,
...>   population_size: 30, max_generations: 100, seed: 99
...> })
iex> {_chrom, f} = result.best
iex> f
8

Config (real encoding)

iex> fitness = fn %Petri.Chromosome.Real{genes: [x, y]} -> -(x*x + y*y) end
iex> result = Petri.run(fitness, %{
...>   encoding: :real, bounds: [{-5.0, 5.0}, {-5.0, 5.0}],
...>   selection: :tournament,
...>   population_size: 50, max_generations: 50, seed: 1
...> })
iex> {_chrom, r2} = result.best
iex> r2 < 0.0
true

Config (permutation encoding)

iex> fitness = fn %Petri.Chromosome.Permutation{genes: g} ->
iex>   # Count adjacent pairs in ascending order
...>   g |> Enum.chunk_every(2, 1, :discard) |> Enum.count(fn [a, b] -> a < b end)
...> end
iex> result = Petri.run(fitness, %{
...>   encoding: :permutation, n: 20,
...>   selection: :tournament,
...>   population_size: 50, max_generations: 100, seed: 1,
...>   crossover: :ox, mutation: :swap
...> })
iex> {_chrom, p_fit} = result.best
iex> p_fit > 10
true