eckity.algorithms.simple_evolution
This module implements the SimpleEvolution class.
1""" 2This module implements the SimpleEvolution class. 3""" 4 5from overrides import overrides 6 7from eckity.algorithms.algorithm import Algorithm 8from eckity.breeders.simple_breeder import SimpleBreeder 9from eckity.evaluators import SimplePopulationEvaluator 10from eckity.random import RNG 11from eckity.termination_checkers import ThresholdFromTargetTerminationChecker 12 13 14class SimpleEvolution(Algorithm): 15 """ 16 Simple case evolutionary algorithm. 17 18 Basic evolutionary algorithm that contains one subpopulation. 19 Does not include and is not meant for multi-objective, co-evolution etc. 20 Such algorithms should be implemented in a new Algorithm subclass. 21 22 Parameters 23 ---------- 24 population: Population 25 The population to be evolved. 26 Contains only one subpopulation in simple case. 27 28 29 statistics: Union[Statistics, List[Statistics]], default=None 30 Provide multiple statistics on the population during the evolution. 31 32 breeder: SimpleBreeder 33 Responsible for applying selection and operators on individuals 34 in each generation. Applies on one subpopulation in simple case. 35 36 population_evaluator: SimplePopulationEvaluator, 37 default=SimplePopulationEvaluator instance 38 Responsible for evaluating each individual's fitness concurrently 39 and returns the best individual of each subpopulation 40 (returns a single individual in simple case). 41 42 max_generation: int, default=1000 43 Maximal number of generations to run the evolutionary process. 44 Note the evolution could end before reaching max_generation, 45 depending on the termination checker. 46 47 events: dict(str, dict(object, function)), default=None 48 Dictionary of events, where each event holds 49 a dictionary of (subscriber, callback method). 50 51 event_names: list of strings, default=None 52 Names of events to publish during the evolution. 53 54 termination_checker: TerminationChecker 55 Checks if the evolution should perform early termination. 56 57 max_workers: int, default=None 58 Maximal number of worker nodes for the Executor object that 59 evaluates the fitness of the individuals. 60 61 random_generator: RNG 62 Random number generator 63 64 random_seed: int, default=current system time 65 Initial random seed for deterministic experiment. 66 67 generation_seed: int, default=None 68 Current generation seed. 69 Useful for resuming a previously paused experiment. 70 71 best_of_run_: Individual, default=None 72 The individual with the best fitness in the entire evolution. 73 74 best_of_gen: Individual, default=None 75 The individual that has the best fitness in the current generation. 76 77 worst_of_gen: Individual, default=None 78 The individual that has the worst fitness in current generation. 79 80 generation_num: int, default=0 81 Current generation number 82 """ 83 84 def __init__( 85 self, 86 population, 87 statistics=None, 88 breeder: SimpleBreeder = SimpleBreeder(), 89 population_evaluator: SimplePopulationEvaluator = None, 90 max_generation=500, 91 events=None, 92 event_names=None, 93 termination_checker=None, 94 executor="thread", 95 max_workers=None, 96 random_generator: RNG = RNG(), 97 random_seed=None, 98 generation_seed=None, 99 best_of_run_=None, 100 best_of_gen=None, 101 worst_of_gen=None, 102 generation_num=0, 103 ): 104 105 if event_names is None: 106 _event_names = [ 107 "before_eval", 108 "after_eval", 109 "before_breeding", 110 "after_breeding", 111 ] 112 else: 113 _event_names = event_names 114 115 if statistics is None: 116 statistics = [] 117 118 if breeder is None: 119 breeder = SimpleBreeder() 120 121 if population_evaluator is None: 122 population_evaluator = SimplePopulationEvaluator() 123 124 if not isinstance(breeder, SimpleBreeder): 125 raise ValueError( 126 f"Expected SimpleBreeder, got {type(breeder)}." 127 ) 128 129 if not isinstance(population_evaluator, SimplePopulationEvaluator): 130 raise ValueError( 131 "Expected SimplePopulationEvaluator, " 132 f"got {type(population_evaluator)}." 133 ) 134 135 super().__init__( 136 population, 137 statistics=statistics, 138 breeder=breeder, 139 population_evaluator=population_evaluator, 140 events=events, 141 event_names=_event_names, 142 executor=executor, 143 max_workers=max_workers, 144 random_generator=random_generator, 145 random_seed=random_seed, 146 generation_seed=generation_seed, 147 termination_checker=termination_checker, 148 generation_num=generation_num, 149 ) 150 151 self.termination_checker = termination_checker 152 self.best_of_run_ = best_of_run_ 153 self.best_of_gen = best_of_gen 154 self.worst_of_gen = worst_of_gen 155 self.max_generation = max_generation 156 157 self.final_generation_ = None 158 159 def initialize(self): 160 """ 161 Initialize the evolutionary algorithm 162 163 Register statistics to `after_generation` event 164 """ 165 super().initialize() 166 for stat in self.statistics: 167 self.register("after_generation", stat.write_statistics) 168 169 @overrides 170 def generation_iteration(self, gen: int) -> bool: 171 """ 172 Performs one iteration of the evolutionary run, 173 for the current generation 174 175 Parameters 176 ---------- 177 gen: 178 current generation number (for example, generation #100) 179 180 Returns 181 ------- 182 None. 183 """ 184 185 # breed population 186 self.breeder.breed(self.population) 187 188 # Evaluate the entire population and get the best individual 189 self.best_of_gen = self.population_evaluator.act(self.population) 190 191 if self.best_of_gen.better_than(self.best_of_run_): 192 self.best_of_run_ = self.best_of_gen 193 194 self.worst_of_gen = self.population.sub_populations[ 195 0 196 ].get_worst_individual() 197 198 def execute(self, **kwargs): 199 """ 200 Compute output using best evolved individual. 201 Use `execute` in a non-sklearn setting. 202 Input keyword arguments that set variable values. 203 For example if `terminal_set=['x', 'y', 1, -1]` 204 then call `execute(x=..., y=...)`. 205 206 Parameters 207 ---------- 208 **kwargs : keyword arguments 209 The input to the program (tree). 210 211 Returns 212 ------- 213 object 214 Output as computed by the best individual of the evolution. 215 216 """ 217 return self.best_of_run_.execute(**kwargs) 218 219 def finish(self): 220 """ 221 Finish the evolutionary run by showing the best individual 222 and printing the best fitness 223 """ 224 super().finish() 225 self.best_of_run_.show() 226 227 def get_individual_evaluator(self): 228 return self.population.sub_populations[0].evaluator 229 230 def get_average_fitness( 231 self, 232 ): # TODO move this function to statistics 233 return self.population.get_average_fitness() 234 235 def event_name_to_data(self, event_name): 236 if event_name == "init": 237 return { 238 "population": self.population, 239 "statistics": self.statistics, 240 "breeder": self.breeder, 241 "termination_checker": self.termination_checker, 242 "max_generation": self.max_generation, 243 "events": self.events, 244 "max_workers": self.max_workers, 245 "generation_num": self.generation_num, 246 } 247 248 # default case 249 return { 250 "population": self.population, 251 "best_of_run_": self.best_of_run_, 252 "best_of_gen": self.best_of_gen, 253 "generation_num": self.generation_num, 254 }
15class SimpleEvolution(Algorithm): 16 """ 17 Simple case evolutionary algorithm. 18 19 Basic evolutionary algorithm that contains one subpopulation. 20 Does not include and is not meant for multi-objective, co-evolution etc. 21 Such algorithms should be implemented in a new Algorithm subclass. 22 23 Parameters 24 ---------- 25 population: Population 26 The population to be evolved. 27 Contains only one subpopulation in simple case. 28 29 30 statistics: Union[Statistics, List[Statistics]], default=None 31 Provide multiple statistics on the population during the evolution. 32 33 breeder: SimpleBreeder 34 Responsible for applying selection and operators on individuals 35 in each generation. Applies on one subpopulation in simple case. 36 37 population_evaluator: SimplePopulationEvaluator, 38 default=SimplePopulationEvaluator instance 39 Responsible for evaluating each individual's fitness concurrently 40 and returns the best individual of each subpopulation 41 (returns a single individual in simple case). 42 43 max_generation: int, default=1000 44 Maximal number of generations to run the evolutionary process. 45 Note the evolution could end before reaching max_generation, 46 depending on the termination checker. 47 48 events: dict(str, dict(object, function)), default=None 49 Dictionary of events, where each event holds 50 a dictionary of (subscriber, callback method). 51 52 event_names: list of strings, default=None 53 Names of events to publish during the evolution. 54 55 termination_checker: TerminationChecker 56 Checks if the evolution should perform early termination. 57 58 max_workers: int, default=None 59 Maximal number of worker nodes for the Executor object that 60 evaluates the fitness of the individuals. 61 62 random_generator: RNG 63 Random number generator 64 65 random_seed: int, default=current system time 66 Initial random seed for deterministic experiment. 67 68 generation_seed: int, default=None 69 Current generation seed. 70 Useful for resuming a previously paused experiment. 71 72 best_of_run_: Individual, default=None 73 The individual with the best fitness in the entire evolution. 74 75 best_of_gen: Individual, default=None 76 The individual that has the best fitness in the current generation. 77 78 worst_of_gen: Individual, default=None 79 The individual that has the worst fitness in current generation. 80 81 generation_num: int, default=0 82 Current generation number 83 """ 84 85 def __init__( 86 self, 87 population, 88 statistics=None, 89 breeder: SimpleBreeder = SimpleBreeder(), 90 population_evaluator: SimplePopulationEvaluator = None, 91 max_generation=500, 92 events=None, 93 event_names=None, 94 termination_checker=None, 95 executor="thread", 96 max_workers=None, 97 random_generator: RNG = RNG(), 98 random_seed=None, 99 generation_seed=None, 100 best_of_run_=None, 101 best_of_gen=None, 102 worst_of_gen=None, 103 generation_num=0, 104 ): 105 106 if event_names is None: 107 _event_names = [ 108 "before_eval", 109 "after_eval", 110 "before_breeding", 111 "after_breeding", 112 ] 113 else: 114 _event_names = event_names 115 116 if statistics is None: 117 statistics = [] 118 119 if breeder is None: 120 breeder = SimpleBreeder() 121 122 if population_evaluator is None: 123 population_evaluator = SimplePopulationEvaluator() 124 125 if not isinstance(breeder, SimpleBreeder): 126 raise ValueError( 127 f"Expected SimpleBreeder, got {type(breeder)}." 128 ) 129 130 if not isinstance(population_evaluator, SimplePopulationEvaluator): 131 raise ValueError( 132 "Expected SimplePopulationEvaluator, " 133 f"got {type(population_evaluator)}." 134 ) 135 136 super().__init__( 137 population, 138 statistics=statistics, 139 breeder=breeder, 140 population_evaluator=population_evaluator, 141 events=events, 142 event_names=_event_names, 143 executor=executor, 144 max_workers=max_workers, 145 random_generator=random_generator, 146 random_seed=random_seed, 147 generation_seed=generation_seed, 148 termination_checker=termination_checker, 149 generation_num=generation_num, 150 ) 151 152 self.termination_checker = termination_checker 153 self.best_of_run_ = best_of_run_ 154 self.best_of_gen = best_of_gen 155 self.worst_of_gen = worst_of_gen 156 self.max_generation = max_generation 157 158 self.final_generation_ = None 159 160 def initialize(self): 161 """ 162 Initialize the evolutionary algorithm 163 164 Register statistics to `after_generation` event 165 """ 166 super().initialize() 167 for stat in self.statistics: 168 self.register("after_generation", stat.write_statistics) 169 170 @overrides 171 def generation_iteration(self, gen: int) -> bool: 172 """ 173 Performs one iteration of the evolutionary run, 174 for the current generation 175 176 Parameters 177 ---------- 178 gen: 179 current generation number (for example, generation #100) 180 181 Returns 182 ------- 183 None. 184 """ 185 186 # breed population 187 self.breeder.breed(self.population) 188 189 # Evaluate the entire population and get the best individual 190 self.best_of_gen = self.population_evaluator.act(self.population) 191 192 if self.best_of_gen.better_than(self.best_of_run_): 193 self.best_of_run_ = self.best_of_gen 194 195 self.worst_of_gen = self.population.sub_populations[ 196 0 197 ].get_worst_individual() 198 199 def execute(self, **kwargs): 200 """ 201 Compute output using best evolved individual. 202 Use `execute` in a non-sklearn setting. 203 Input keyword arguments that set variable values. 204 For example if `terminal_set=['x', 'y', 1, -1]` 205 then call `execute(x=..., y=...)`. 206 207 Parameters 208 ---------- 209 **kwargs : keyword arguments 210 The input to the program (tree). 211 212 Returns 213 ------- 214 object 215 Output as computed by the best individual of the evolution. 216 217 """ 218 return self.best_of_run_.execute(**kwargs) 219 220 def finish(self): 221 """ 222 Finish the evolutionary run by showing the best individual 223 and printing the best fitness 224 """ 225 super().finish() 226 self.best_of_run_.show() 227 228 def get_individual_evaluator(self): 229 return self.population.sub_populations[0].evaluator 230 231 def get_average_fitness( 232 self, 233 ): # TODO move this function to statistics 234 return self.population.get_average_fitness() 235 236 def event_name_to_data(self, event_name): 237 if event_name == "init": 238 return { 239 "population": self.population, 240 "statistics": self.statistics, 241 "breeder": self.breeder, 242 "termination_checker": self.termination_checker, 243 "max_generation": self.max_generation, 244 "events": self.events, 245 "max_workers": self.max_workers, 246 "generation_num": self.generation_num, 247 } 248 249 # default case 250 return { 251 "population": self.population, 252 "best_of_run_": self.best_of_run_, 253 "best_of_gen": self.best_of_gen, 254 "generation_num": self.generation_num, 255 }
Simple case evolutionary algorithm.
Basic evolutionary algorithm that contains one subpopulation. Does not include and is not meant for multi-objective, co-evolution etc. Such algorithms should be implemented in a new Algorithm subclass.
Parameters
- population (Population): The population to be evolved. Contains only one subpopulation in simple case.
- statistics (Union[Statistics, List[Statistics]], default=None): Provide multiple statistics on the population during the evolution.
- breeder (SimpleBreeder): Responsible for applying selection and operators on individuals in each generation. Applies on one subpopulation in simple case.
- population_evaluator (SimplePopulationEvaluator,): default=SimplePopulationEvaluator instance Responsible for evaluating each individual's fitness concurrently and returns the best individual of each subpopulation (returns a single individual in simple case).
- max_generation (int, default=1000): Maximal number of generations to run the evolutionary process. Note the evolution could end before reaching max_generation, depending on the termination checker.
- events (dict(str, dict(object, function)), default=None): Dictionary of events, where each event holds a dictionary of (subscriber, callback method).
- event_names (list of strings, default=None): Names of events to publish during the evolution.
- termination_checker (TerminationChecker): Checks if the evolution should perform early termination.
- max_workers (int, default=None): Maximal number of worker nodes for the Executor object that evaluates the fitness of the individuals.
- random_generator (RNG): Random number generator
- random_seed (int, default=current system time): Initial random seed for deterministic experiment.
- generation_seed (int, default=None): Current generation seed. Useful for resuming a previously paused experiment.
- best_of_run_ (Individual, default=None): The individual with the best fitness in the entire evolution.
- best_of_gen (Individual, default=None): The individual that has the best fitness in the current generation.
- worst_of_gen (Individual, default=None): The individual that has the worst fitness in current generation.
- generation_num (int, default=0): Current generation number
85 def __init__( 86 self, 87 population, 88 statistics=None, 89 breeder: SimpleBreeder = SimpleBreeder(), 90 population_evaluator: SimplePopulationEvaluator = None, 91 max_generation=500, 92 events=None, 93 event_names=None, 94 termination_checker=None, 95 executor="thread", 96 max_workers=None, 97 random_generator: RNG = RNG(), 98 random_seed=None, 99 generation_seed=None, 100 best_of_run_=None, 101 best_of_gen=None, 102 worst_of_gen=None, 103 generation_num=0, 104 ): 105 106 if event_names is None: 107 _event_names = [ 108 "before_eval", 109 "after_eval", 110 "before_breeding", 111 "after_breeding", 112 ] 113 else: 114 _event_names = event_names 115 116 if statistics is None: 117 statistics = [] 118 119 if breeder is None: 120 breeder = SimpleBreeder() 121 122 if population_evaluator is None: 123 population_evaluator = SimplePopulationEvaluator() 124 125 if not isinstance(breeder, SimpleBreeder): 126 raise ValueError( 127 f"Expected SimpleBreeder, got {type(breeder)}." 128 ) 129 130 if not isinstance(population_evaluator, SimplePopulationEvaluator): 131 raise ValueError( 132 "Expected SimplePopulationEvaluator, " 133 f"got {type(population_evaluator)}." 134 ) 135 136 super().__init__( 137 population, 138 statistics=statistics, 139 breeder=breeder, 140 population_evaluator=population_evaluator, 141 events=events, 142 event_names=_event_names, 143 executor=executor, 144 max_workers=max_workers, 145 random_generator=random_generator, 146 random_seed=random_seed, 147 generation_seed=generation_seed, 148 termination_checker=termination_checker, 149 generation_num=generation_num, 150 ) 151 152 self.termination_checker = termination_checker 153 self.best_of_run_ = best_of_run_ 154 self.best_of_gen = best_of_gen 155 self.worst_of_gen = worst_of_gen 156 self.max_generation = max_generation 157 158 self.final_generation_ = None
160 def initialize(self): 161 """ 162 Initialize the evolutionary algorithm 163 164 Register statistics to `after_generation` event 165 """ 166 super().initialize() 167 for stat in self.statistics: 168 self.register("after_generation", stat.write_statistics)
Initialize the evolutionary algorithm
Register statistics to after_generation event
170 @overrides 171 def generation_iteration(self, gen: int) -> bool: 172 """ 173 Performs one iteration of the evolutionary run, 174 for the current generation 175 176 Parameters 177 ---------- 178 gen: 179 current generation number (for example, generation #100) 180 181 Returns 182 ------- 183 None. 184 """ 185 186 # breed population 187 self.breeder.breed(self.population) 188 189 # Evaluate the entire population and get the best individual 190 self.best_of_gen = self.population_evaluator.act(self.population) 191 192 if self.best_of_gen.better_than(self.best_of_run_): 193 self.best_of_run_ = self.best_of_gen 194 195 self.worst_of_gen = self.population.sub_populations[ 196 0 197 ].get_worst_individual()
Performs one iteration of the evolutionary run, for the current generation
Parameters
- gen:: current generation number (for example, generation #100)
Returns
- None.
199 def execute(self, **kwargs): 200 """ 201 Compute output using best evolved individual. 202 Use `execute` in a non-sklearn setting. 203 Input keyword arguments that set variable values. 204 For example if `terminal_set=['x', 'y', 1, -1]` 205 then call `execute(x=..., y=...)`. 206 207 Parameters 208 ---------- 209 **kwargs : keyword arguments 210 The input to the program (tree). 211 212 Returns 213 ------- 214 object 215 Output as computed by the best individual of the evolution. 216 217 """ 218 return self.best_of_run_.execute(**kwargs)
Compute output using best evolved individual.
Use execute in a non-sklearn setting.
Input keyword arguments that set variable values.
For example if terminal_set=['x', 'y', 1, -1]
then call execute(x=..., y=...).
Parameters
- **kwargs (keyword arguments): The input to the program (tree).
Returns
- object: Output as computed by the best individual of the evolution.
220 def finish(self): 221 """ 222 Finish the evolutionary run by showing the best individual 223 and printing the best fitness 224 """ 225 super().finish() 226 self.best_of_run_.show()
Finish the evolutionary run by showing the best individual and printing the best fitness
236 def event_name_to_data(self, event_name): 237 if event_name == "init": 238 return { 239 "population": self.population, 240 "statistics": self.statistics, 241 "breeder": self.breeder, 242 "termination_checker": self.termination_checker, 243 "max_generation": self.max_generation, 244 "events": self.events, 245 "max_workers": self.max_workers, 246 "generation_num": self.generation_num, 247 } 248 249 # default case 250 return { 251 "population": self.population, 252 "best_of_run_": self.best_of_run_, 253 "best_of_gen": self.best_of_gen, 254 "generation_num": self.generation_num, 255 }
Convert event name to relevant data of the Algorithm for the event
Parameters
- event_name (string): name of the event that is happening
Returns
- Dict[str, object]: Algorithm data regarding the given event