def _should_do_rule_discovery_in_action_set(self, action_set):
mean_time_stamp_in_action_set = calc_summary_stat(
action_set, "mean", "time_stamp")
time_since_last_rule_discovery = self._time_step - \
mean_time_stamp_in_action_set
return time_since_last_rule_discovery > \
get_hyperparam("theta_ga") and self._did_explore
def __call__(self, action_set, payoff, situation):
niche_min_error = calc_summary_stat(action_set, "min", "error")
for classifier in action_set:
classifier.experience += 1
payoff_diff = payoff - classifier.get_prediction(situation)
self._update_weight_vec(classifier, payoff_diff, situation)
self._update_niche_min_error(classifier, niche_min_error)
self._update_prediction_error(classifier, payoff_diff)
update_action_set_size(classifier, action_set)
def _select_for_deletion(self, population):
"""First loop (selecting classifier to delete) of
DELETE FROM POPULATION function from 'An Algorithmic Description of
XCS' (Butz and Wilson, 2002).
This method should only ever be called if the population
is past its microclassifier capacity."""
mean_fitness_in_pop = calc_summary_stat(population, "mean", "fitness")
votes = [
self._calc_deletion_vote(classifier, mean_fitness_in_pop)
for classifier in population
]
vote_sum = sum(votes)
choice_point = get_rng().rand() * vote_sum
vote_sum = 0
for (classifier, vote) in zip(population, votes):
vote_sum += vote
if vote_sum > choice_point:
return classifier
def test_sample_training_run(self):
# 6-mux
env = make_discrete_mux_env(num_address_bits=2,
shuffle_dataset=True,
reward_correct=1000,
reward_incorrect=0)
rule_repr = DiscreteRuleRepr()
alg_hyperparams = {
"N": 400,
"beta": 0.2,
"alpha": 0.1,
"epsilon_nought": 0.01,
"nu": 5,
"gamma": 0.71,
"theta_ga": 12,
"chi": 0.8,
"mu": 0.04,
"theta_del": 20,
"delta": 0.1,
"theta_sub": 20,
"p_wildcard": 0.33,
"prediction_I": 1e-3,
"epsilon_I": 1e-3,
"fitness_I": 1e-3,
"p_explore": 0.5,
"theta_mna": len(env.action_set),
"do_ga_subsumption": True,
"do_as_subsumption": True,
"seed": 0
}
alg = make_canonical_xcs(env, rule_repr, alg_hyperparams)
monitor_items = [
MonitorItem("num_micros",
lambda experiment: experiment.population.num_micros),
MonitorItem("num_macros",
lambda experiment: experiment.population.num_macros),
MonitorItem(
"performance", lambda experiment: experiment.calc_performance(
strat="accuracy")),
MonitorItem(
"mean_error", lambda experiment: calc_summary_stat(
experiment.population, "mean", "error")),
MonitorItem(
"max_fitness", lambda experiment: calc_summary_stat(
experiment.population, "max", "fitness")),
MonitorItem(
"deletion", lambda experiment: experiment.population.
operations_record["deletion"]),
MonitorItem(
"covering", lambda experiment: experiment.population.
operations_record["covering"]),
MonitorItem(
"as_subsumption", lambda experiment: experiment.population.
operations_record["as_subsumption"]),
MonitorItem(
"ga_subsumption", lambda experiment: experiment.population.
operations_record["ga_subsumption"]),
MonitorItem(
"discovery", lambda experiment: experiment.population.
operations_record["discovery"]),
MonitorItem(
"absorption", lambda experiment: experiment.population.
operations_record["absorption"])
]
experiment = Experiment(env,
alg,
num_training_epochs=10,
monitor_items=monitor_items)
experiment.run()
print(experiment._monitor.query())