eckity.multi_objective_evolution.nsga2_evolution

  1from time import time
  2from overrides import overrides
  3
  4from eckity.algorithms.algorithm import Algorithm
  5from eckity.breeders.simple_breeder import SimpleBreeder
  6from eckity.evaluators.simple_population_evaluator import SimplePopulationEvaluator
  7from eckity.multi_objective_evolution.nsga2_front_sorting import NSGA2FrontSorting
  8
  9from eckity.termination_checkers.threshold_from_target_termination_checker \
 10	import ThresholdFromTargetTerminationChecker
 11
 12
 13class NSGA2Evolution(Algorithm):
 14	def __init__(self,
 15				 population,
 16				 statistics=None,
 17				 breeder=SimpleBreeder(),
 18				 population_evaluator=SimplePopulationEvaluator(),
 19				 NSGA2FrontSorting=NSGA2FrontSorting(),
 20				 max_generation=500,
 21				 events=None,
 22				 event_names=None,
 23				 termination_checker=ThresholdFromTargetTerminationChecker(threshold=0),
 24				 max_workers=None,
 25				 random_generator=None,
 26				 random_seed=time(),
 27				 generation_seed=None,
 28				 best_of_run_=None,
 29				 best_of_run_evaluator=None,
 30				 best_of_gen=None,
 31				 worst_of_gen=None,
 32				 generation_num=0):
 33
 34		if event_names is None:
 35			_event_names = ['before_eval', 'after_eval', 'before_breeding', 'after_breeding']
 36		else:
 37			_event_names = event_names
 38
 39		if statistics is None:
 40			statistics = []
 41
 42		super().__init__(population, statistics=statistics, breeder=breeder, population_evaluator=population_evaluator,
 43						 events=events, event_names=_event_names, max_workers=max_workers,
 44						 random_generator=random_generator, random_seed=random_seed, generation_seed=generation_seed,
 45						 termination_checker=termination_checker, generation_num=generation_num)
 46
 47		self.termination_checker = termination_checker
 48		self.best_of_run_ = best_of_run_
 49		self.best_of_run_evaluator = best_of_run_evaluator
 50		self.best_of_gen = best_of_gen
 51		self.worst_of_gen = worst_of_gen
 52		self.max_generation = max_generation
 53
 54		self.final_generation_ = None
 55		self.NSGA2FrontSorting = NSGA2FrontSorting
 56
 57	#
 58	@overrides
 59	def generation_iteration(self, gen):
 60		"""
 61		Performs one iteration of the evolutionary run, for the current generation
 62
 63		Parameters
 64		----------
 65		gen:
 66			current generation number (for example, generation #100)
 67
 68		Returns
 69		-------
 70		None.
 71		"""
 72
 73		# breed population
 74		self.breeder.breed(self.population)
 75		# Evaluate the entire population and get the best individual
 76
 77		self.population_evaluator.act(self.population)
 78
 79		self.NSGA2FrontSorting.select_for_population(self.population)
 80		self.best_of_gen = self._get_pareto_fronts()
 81
 82	def _get_pareto_fronts(self):
 83		'''
 84		Returns: the pareto front for each sub_population
 85		-------
 86		'''
 87		pareto_fronts = []
 88		for sub_pop in self.population.sub_populations:
 89			pareto_fronts.append([ind for ind in sub_pop.individuals if ind.fitness.front_rank == 1])
 90		return pareto_fronts
 91
 92	def initialize(self):
 93		"""
 94		Initialize the evolutionary algorithm
 95
 96		Register statistics to `after_generation` event
 97		"""
 98		super().initialize()
 99		for stat in self.statistics:
100			self.register('after_generation', stat.write_statistics)
101
102	def execute(self, **kwargs):
103		"""
104		Compute output using best evolved individual.
105		Use `execute` in a non-sklearn setting.
106		Input keyword arguments that set variable values.
107		For example if `terminal_set=['x', 'y', 1, -1]` then call `execute(x=..., y=...)`.
108
109		Parameters
110		----------
111		**kwargs : keyword arguments
112			The input to the program (tree).
113
114		Returns
115		-------
116		object
117			Output as computed by the best individual of the evolutionary run.
118
119		"""
120		return self.best_of_run_.execute(**kwargs)
121
122	@overrides
123	def finish(self):
124		"""
125		Finish the evolutionary run by showing the best individual and printing the best fitness
126		"""
127		self.best_of_run_.show()
128		print(self.best_of_run_.get_pure_fitness())
129
130	def event_name_to_data(self, event_name):
131		if event_name == "init":
132			return {
133				"population": self.population,
134				"statistics": self.statistics,
135				"breeder": self.breeder,
136				"termination_checker": self.termination_checker,
137				"max_generation": self.max_generation,
138				"events": self.events,
139				"max_workers": self.max_workers
140			}
141
142		# default case
143		return {
144			"population": self.population,
145			"best_of_run_": self.best_of_run_,
146			"best_of_gen": self.best_of_gen,
147			"generation_num": self.generation_num
148		}