Merge pull request #247 from alcides/problem_with_target

Problems now store the target value

docs/source/stopping.md

@@ -14,20 +14,6 @@ Search algorithms can be defined with different budgets (subclasses of `genetice
 .. autoapiclass:: geneticengine.evaluation.budget.EvaluationBudget
 ```
 
-## Target Fitness
-
-For single-objective:
-
-```{eval-rst}
-.. autoapiclass:: geneticengine.evaluation.budget.TargetFitness
-```
-
-For multi-objective:
-
-```{eval-rst}
-.. autoapiclass:: geneticengine.evaluation.budget.TargetMultiFitness
-```
-
 ## Budget Combinators
 
 ## AnyOf

examples/benchmarks/benchmark.py

@@ -21,7 +21,7 @@ def example_run(b: Benchmark):
     problem = b.get_problem()
     grammar = b.get_grammar()
     random = NativeRandomSource(1)
-    s = GeneticProgramming(
+    alg = GeneticProgramming(
         problem=problem,
         budget=TimeBudget(5),
         representation=TreeBasedRepresentation(
@@ -31,7 +31,7 @@ def example_run(b: Benchmark):
         random=random,
         tracker=ProgressTracker(problem, recorders=[PopulationRecorder()]),
     )
-    best = s.search()
+    best = alg.search()[0]
     print(
         f"Fitness of {best.get_fitness(problem)} by genotype: {best.genotype} with phenotype: {best.get_phenotype()}",
     )

examples/benchmarks/benchmark_suite.py

@@ -59,7 +59,8 @@
                 recorders=[PopulationRecorder()],
             )
             instance = m(grammar, problem, random, budget, tracker)
-            best = instance.search()
+            bests = instance.search()
+            best = bests[0]
             print(
                 f"Fitness of {best.get_fitness(problem)} by genotype: {best.genotype} with phenotype: {best.get_phenotype()}",
             )

examples/benchmarks/classification.py

@@ -16,7 +16,6 @@
 from geml.grammars.ruleset_classification import make_grammar
 
 
-
 class ClassificationBenchmark(Benchmark):
 
     def __init__(self, X, y, feature_names):
@@ -34,7 +33,7 @@ def fitness_function(ruleset):
             except ValueError:
                 return -1
 
-        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function)
+        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function, target=1)
 
     def setup_grammar(self, y, feature_names):
         # Grammar

examples/benchmarks/classification_lexicase.py

@@ -5,7 +5,7 @@
 from examples.benchmarks.benchmark import Benchmark, example_run
 from examples.benchmarks.datasets import get_banknote
 from geneticengine.grammar.grammar import extract_grammar, Grammar
-from geneticengine.problems import Problem, MultiObjectiveProblem
+from geneticengine.problems import MultiObjectiveProblem, Problem
 from geml.common import forward_dataset
 from geml.grammars.ruleset_classification import make_grammar
 
@@ -24,7 +24,12 @@ def fitness_function_lexicase(ruleset):
             except ValueError:
                 return [0 for _ in y]
 
-        self.problem = MultiObjectiveProblem(minimize=False, fitness_function=fitness_function_lexicase)
+        n_objectives = len(y)
+        self.problem = MultiObjectiveProblem(
+            fitness_function=fitness_function_lexicase,
+            minimize=[False for _ in range(n_objectives)],
+            target=[1 for _ in range(n_objectives)],
+        )
 
     def setup_grammar(self, y, feature_names):
         options = [int(v) for v in np.unique(y)]

examples/benchmarks/game_of_life_vectorial.py

@@ -185,7 +185,7 @@ def fitness_function(i: Condition):
             ypred = [_clf(line) for line in np.rollaxis(X, 0)]
             return f1_score(y, ypred)
 
-        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function)
+        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function, target=1)
 
     def setup_grammar(self):
         self.grammar = extract_grammar(

examples/benchmarks/regression.py

@@ -33,7 +33,7 @@ def fitness_function(ruleset):
             except ValueError:
                 return -1
 
-        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function)
+        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function, target=0)
 
     def setup_grammar(self, feature_names):
         # Grammar

examples/benchmarks/regression_lexicase.py

@@ -61,7 +61,12 @@ def lexicase_fitness_function(n: Expression):
 
             return grouped_errors
 
-        self.problem = MultiObjectiveProblem(minimize=True, fitness_function=lexicase_fitness_function)
+        n_objectives = len(y)
+        self.problem = MultiObjectiveProblem(
+            fitness_function=lexicase_fitness_function,
+            minimize=[True for _ in range(n_objectives)],
+            target=[0 for _ in range(n_objectives)],
+        )
 
     def setup_grammar(self, feature_names):
         # Grammar

examples/benchmarks/santafe.py

@@ -174,7 +174,8 @@ def setup_problem(self, map):
         def fitness_function(i):
             return simulate(i, map)
 
-        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function)
+        target_food = map.count("#")
+        self.problem = SingleObjectiveProblem(minimize=False, fitness_function=fitness_function, target=target_food)
 
     def setup_grammar(self):
         self.grammar = extract_grammar([ActionBlock, Action, IfFood, Move, Right, Left], ActionMain)

examples/benchmarks/string_match.py

@@ -1,8 +1,6 @@
 from __future__ import annotations
 
 
-
-
 from examples.benchmarks.benchmark import Benchmark, example_run
 from geml.grammars.letter import Char, Consonant, LetterString, String, Vowel
 from geneticengine.grammar.grammar import extract_grammar
@@ -11,7 +9,6 @@
 from geneticengine.problems import SingleObjectiveProblem
 
 
-
 class StringMatchBenchmark(Benchmark):
     def __init__(self, target: str = "Hellow world!"):
         self.setup_problem(target)
@@ -33,7 +30,7 @@ def fitness_function(individual):
                     fitness -= 1 / (1 + (abs(ord(t_p) - ord(g_p))))
             return fitness
 
-        self.problem = SingleObjectiveProblem(minimize=True, fitness_function=fitness_function)
+        self.problem = SingleObjectiveProblem(minimize=True, fitness_function=fitness_function, target=0)
 
     def setup_grammar(self):
         self.grammar = extract_grammar([LetterString, Char, Vowel, Consonant], String)

examples/benchmarks/vectorialgp.py

@@ -3112,7 +3112,7 @@ def regressor(l):
                 y_pred,
             )
 
-        self.problem = SingleObjectiveProblem(minimize=True, fitness_function=fitness_function)
+        self.problem = SingleObjectiveProblem(minimize=True, fitness_function=fitness_function, target=0)
 
     def setup_grammar(self):
         """See fitness/vectorialgp.py for docs.

examples/binary.py

@@ -3,7 +3,7 @@
 from dataclasses import dataclass
 from typing import Annotated
 
-from geneticengine.evaluation.budget import AnyOf, TargetFitness, TimeBudget
+from geneticengine.evaluation.budget import TimeBudget
 from geneticengine.problems import SingleObjectiveProblem
 from geneticengine.random.sources import NativeRandomSource, RandomSource
 
@@ -87,8 +87,8 @@ def crossover(
     repr = BinaryListTreeBasedRepresentation(random=r, grammar=g, max_depth=4)
 
     gp = GeneticProgramming(
-        problem=SingleObjectiveProblem(fitness_function=fitness, minimize=False),
-        budget=AnyOf(TargetFitness(1), TimeBudget(3)),
+        problem=SingleObjectiveProblem(fitness_function=fitness, minimize=False, target=1),
+        budget=TimeBudget(3),
         representation=repr,
         random=r,
         population_size=10,

examples/classification_unknown_length_objectives.py

@@ -7,7 +7,8 @@
 import pandas as pd
 from sklearn.model_selection import train_test_split
 
-from geml.simplegp import SimpleGP
+from geneticengine.algorithms.gp.gp import GeneticProgramming
+from geneticengine.evaluation.budget import EvaluationBudget
 from geneticengine.grammar.decorators import abstract
 from geneticengine.grammar.grammar import extract_grammar
 from geneticengine.grammar.grammar import Grammar
@@ -20,6 +21,10 @@
 from geml.grammars.sgp import Var
 from geneticengine.grammar.metahandlers.floats import FloatList
 from geneticengine.grammar.metahandlers.vars import VarRange
+from geneticengine.problems import LazyMultiObjectiveProblem
+from geneticengine.random.sources import NativeRandomSource
+from geneticengine.representations.tree.initializations import MaxDepthDecider
+from geneticengine.representations.tree.treebased import TreeBasedRepresentation
 
 # ===================================
 # This is an example of normal classification using normal GP,
@@ -179,22 +184,16 @@ def get_grammar(self) -> Grammar:
         )
 
     def main(self, **args):
-        g = self.get_grammar()
-
-        alg = SimpleGP(
-            grammar=g,
-            minimize=[],
-            fitness_function=fitness_function_lexicase,
-            crossover_probability=0.75,
-            mutation_probability=0.01,
-            max_evaluations=10000,
-            max_depth=15,
+        grammar = self.get_grammar()
+        random = NativeRandomSource(0)
+        problem = LazyMultiObjectiveProblem(fitness_function_lexicase, minimize=False, target=1)
+        alg = GeneticProgramming(
+            problem=problem,
+            representation=TreeBasedRepresentation(grammar, MaxDepthDecider(random, grammar, 15)),
+            budget=EvaluationBudget(1000),
             population_size=50,
-            selection_method=("lexicase",),
-            elitism=5,
-            **args,
         )
-        best = alg.search()
+        best = alg.search()[0]
         print(
             f"Fitness of {best.get_fitness(alg.get_problem())} by genotype: {best.genotype} with phenotype: {best.get_phenotype()}",
         )

examples/coevolution_dna.py

@@ -3,7 +3,7 @@
 from typing import Annotated
 
 from geneticengine.algorithms.gp.cooperativegp import CooperativeGP
-from geneticengine.evaluation.budget import AnyOf, EvaluationBudget, TargetFitness
+from geneticengine.evaluation.budget import EvaluationBudget
 from geneticengine.grammar.grammar import extract_grammar
 from geneticengine.grammar.metahandlers.lists import ListSizeBetween
 from geneticengine.grammar.metahandlers.strings import StringSizeBetween
@@ -47,8 +47,7 @@ def dataset_fitness_function(d: Dataset):
 
 class Explanation(ABC):
     @abstractmethod
-    def eval(self, line: str) -> bool:
-        ...
+    def eval(self, line: str) -> bool: ...
 
 
 @dataclass
@@ -118,10 +117,10 @@ def eval_explanation_on_dataset(d: Dataset, e: Explanation) -> float:
     population2_size=500,
     coevolutions=10,
     kwargs1={
-        "budget": AnyOf(TargetFitness(55), EvaluationBudget(10000)),
+        "budget": EvaluationBudget(10000),
     },
     kwargs2={
-        "budget": AnyOf(TargetFitness(100), EvaluationBudget(10000)),
+        "budget": EvaluationBudget(10000),
     },
 )
 x: tuple[Dataset, Explanation] = alg.search()

examples/example.py

@@ -56,17 +56,17 @@ def target(x):
 
 
 alg_gp = GeneticProgramming(problem=problem, budget=budget, representation=representation)
-ind = alg_gp.search()
+ind = alg_gp.search()[0]
 print("\n======\nGP\n======\n")
 print(f"{ind.get_fitness(problem)} - {ind}")
 
 
 alg_hc = HC(problem=problem, budget=budget, representation=representation)
-ind = alg_hc.search()
+ind = alg_hc.search()[0]
 print("\n======\nHC\n======\n")
 print(f"{ind.get_fitness(problem)} - {ind}")
 
 alg_rs = RandomSearch(problem=problem, budget=budget, representation=representation)
-ind = alg_rs.search()
+ind = alg_rs.search()[0]
 print("\n======\nRS\n======\n")
 print(f"{ind.get_fitness(problem)} - {ind}")

examples/extra_columns_example.py

@@ -59,5 +59,5 @@ def fitness_test(p) -> float:
     seed=SEED,
 )
 
-ind = gp.search()
+ind = gp.search()[0]
 print(simplify(ind.get_phenotype()))

examples/genetic_example/genetic_example.py

@@ -17,13 +17,11 @@
 
 
 class DecisionNode(ABC):
-    def evaluate(self, X):
-        ...
+    def evaluate(self, X): ...
 
 
 class IntNode(ABC):
-    def evaluate(self, X):
-        ...
+    def evaluate(self, X): ...
 
 
 @dataclass
@@ -101,7 +99,7 @@ def fitness_function(d: DecisionNode):
         population_size=10,
         max_evaluations=10000,
     )
-    ind = alg.search()
+    ind = alg.search()[0]
     print(ind)
 
 

examples/logging_example.py

@@ -67,6 +67,6 @@ def target(x):
 
 
 alg_gp = GeneticProgramming(problem=problem, budget=budget, representation=representation, population_size=10)
-ind = alg_gp.search()
+ind = alg_gp.search()[0]
 print("\n======\nGP\n======\n")
 print(f"{ind.get_fitness(problem)} - {ind}")

examples/multi_target_lexicase.py

@@ -164,7 +164,7 @@ def main(self, **args):
             elitism=0,
             **args,
         )
-        best = alg.search()
+        best = alg.search()[0]
         print(
             f"Fitness of {best.get_fitness(alg.get_problem())} by genotype: {best.genotype} with phenotype: {best.get_phenotype()}",
         )

examples/pcfg_example.py

@@ -48,7 +48,7 @@ def fitness_function(x) -> float:
         population_size=1000,
         max_evaluations=10 * 1000,
     )
-    ind = alg.search()
+    ind = alg.search()[0]
 
     x: TreeNode = ind.get_phenotype()
     print(x)

examples/progsys/Median.py

@@ -109,7 +109,7 @@ def evolve(g, seed, mode, representation="treebased"):
         population_size=50,
         crossover_probability=0.9,
     )
-    ind = alg.search()
+    ind = alg.search()[0]
     return ind.get_phenotype(), ind.get_fitness(alg.get_problem()), g
 
 

examples/progsys/Number_IO.py

@@ -60,7 +60,7 @@ def evolve(g, seed, mode, representation="treebased"):
         population_size=50,
         crossover_probability=0.9,
     )
-    ind = alg.search()
+    ind = alg.search()[0]
     return ind.get_phenotype(), ind.get_fitness(alg.get_problem()), g
 
 

examples/progsys/Smallest.py

@@ -135,7 +135,7 @@ def evolve(g, seed, mode, representation="treebased"):
         population_size=50,
         crossover_probability=0.9,
     )
-    ind = alg.search()
+    ind = alg.search()[0]
     return ind.get_phenotype(), ind.get_fitness(alg.get_problem()), g
 
 

examples/progsys/Sum_of_Squares.py

@@ -109,7 +109,7 @@ def evolve(g, seed, mode, representation="treebased"):
         population_size=50,
         crossover_probability=0.9,
     )
-    ind = alg.search()
+    ind = alg.search()[0]
     return ind.get_phenotype(), ind.get_fitness(alg.get_problem()), g
 
 

examples/progsys/Vector_Average.py

@@ -111,7 +111,7 @@ def evolve(g, seed, mode, representation="treebased"):
         population_size=50,
         crossover_probability=0.9,
     )
-    ind = alg.search()
+    ind = alg.search()[0]
     return ind.get_phenotype(), ind.fitness, g