def create_majority_step_swarm():
swarm = StepSwarm(
model=lambda x: DiscreteUniform(env=x),
env=lambda: ParallelEnvironment(lambda: DiscreteEnv(ClassicControl())),
reward_limit=10,
n_walkers=100,
max_epochs=20,
step_epochs=25,
)
return swarm
def create_step_to_best():
swarm = StepToBest(
model=lambda x: DiscreteUniform(env=x),
env=lambda: ParallelEnvironment(lambda: DiscreteEnv(ClassicControl())),
reward_limit=16,
n_walkers=100,
max_epochs=5,
step_epochs=25,
)
return swarm
def create_cartpole_swarm():
swarm = Swarm(
model=lambda x: DiscreteUniform(env=x),
walkers=Walkers,
env=lambda: DiscreteEnv(ClassicControl()),
n_walkers=20,
max_iters=200,
prune_tree=True,
reward_scale=2,
)
return swarm
def create_cartpole_swarm():
swarm = Swarm(
model=lambda x: DiscreteUniform(env=x),
walkers=Walkers,
env=lambda: DiscreteEnv(ClassicControl()),
reward_limit=121,
n_walkers=150,
max_epochs=300,
reward_scale=2,
)
return swarm
def create_follow_best_step_swarm():
swarm = StepSwarm(
root_model=FollowBestModel,
model=lambda x: DiscreteUniform(env=x),
env=lambda: ParallelEnvironment(lambda: DiscreteEnv(ClassicControl())),
reward_limit=15,
n_walkers=100,
max_epochs=15,
step_epochs=25,
)
return swarm
def swarm_with_tree():
swarm = Swarm(
model=lambda x: DiscreteUniform(env=x),
env=lambda: DiscreteEnv(ClassicControl()),
reward_limit=200,
n_walkers=150,
max_epochs=300,
reward_scale=2,
tree=HistoryTree,
prune_tree=True,
)
return swarm
def create_cartpole_swarm():
from fragile.core import DiscreteEnv, DiscreteUniform, Swarm
from plangym.minimal import ClassicControl
swarm = Swarm(
model=lambda x: DiscreteUniform(env=x),
env=lambda: DiscreteEnv(ClassicControl()),
reward_limit=51,
n_walkers=50,
max_epochs=100,
reward_scale=2,
)
return swarm