`evorl.ec.optimizers.vanilla_ga`¶

Module Contents¶

Classes¶

`VanillaGA`	Vanilla Genetic Algorithm.
`VanillaGAState`	State of the VanillaGA.

API¶

class evorl.ec.optimizers.vanilla_ga.VanillaGA[source]¶

Bases: evorl.ec.optimizers.ec_optimizer.EvoOptimizer

Vanilla Genetic Algorithm.

The Genetic Algorithm used in the original ERL. Paper: Evolution-Guided Policy Gradient in Reinforcement Learning

ask(state: evorl.ec.optimizers.vanilla_ga.VanillaGAState) → tuple[chex.ArrayTree, evorl.ec.optimizers.vanilla_ga.VanillaGAState][source]¶

enable_crossover: bool¶: True

init(pop: chex.ArrayTree, key: chex.PRNGKey) → evorl.ec.optimizers.vanilla_ga.VanillaGAState[source]¶

matrix_num_mutation_frac: float¶: 0.01

mut_strength: float¶: 0.1

num_crossover_frac: float¶: 0.1

num_elites: int¶: None

pop_size: int¶: None

tell(state: evorl.ec.optimizers.vanilla_ga.VanillaGAState, fitnesses: chex.Array) → tuple[evorl.types.PyTreeDict, evorl.ec.optimizers.vanilla_ga.VanillaGAState][source]¶

tournament_size: int¶: 2

vector_num_mutation_frac: float¶: 0.0

weight_max_magnitude: float¶: 1000000.0

class evorl.ec.optimizers.vanilla_ga.VanillaGAState[source]¶

Bases: evorl.types.PyTreeData

State of the VanillaGA.

key: chex.PRNGKey¶: None

pop: chex.ArrayTree¶: None