xdcxiang-cell · xdcxiang-cell · Dec 10, 2025
diff --git a/README.md b/README.md
@@ -292,6 +292,79 @@ If you used PyGAD, please consider adding a citation to the following paper abou
 }
 ```
 
+# New Features
+
+## Environment Switching
+
+PyGAD now supports automatic environment switching to adapt the genetic algorithm parameters based on the generation number:
+
+1.  **ENV_STABLE**: Default environment with standard parameters (crossover_probability=0.8, mutation_probability=0.1, parent_selection_type="sss")
+2.  **ENV_FAST**: Optimized for faster convergence (crossover_probability × 1.2, mutation_probability × 0.5, parent_selection_type="tournament")
+3.  **ENV_DIVERSE**: Optimized for maintaining population diversity (crossover_probability × 0.7, mutation_probability × 2, parent_selection_type="tournament_nsga2")
+
+By default, the environment cycles every 10 generations: STABLE → FAST → DIVERSE → STABLE...
+
+### Example Usage
+```python
+import pygad
+import numpy as np
+
+def fitness_func(ga_instance, solution, solution_idx):
+    return np.sum(solution)
+
+# Create GA instance with environment switching enabled
+ga_instance = pygad.GA(
+    num_generations=30,
+    num_parents_mating=5,
+    fitness_func=fitness_func,
+    sol_per_pop=10,
+    num_genes=5
+)
+
+# Run the GA with automatic environment switching
+ga_instance.run()
+```
+
+## Pareto Front Output
+
+For multi-objective optimization problems, PyGAD now automatically calculates and outputs the Pareto front after each generation. The Pareto front represents the set of non-dominated solutions where no solution can be improved in one objective without worsening at least one other objective.
+
+### Key Features
+1.  **Automatic Calculation**: The Pareto front is automatically calculated after each generation for multi-objective fitness functions
+2.  **Serialization**: The Pareto front is included in the generation output dictionary
+3.  **Visualization**: Use `plot_pareto_front_curve()` method to visualize the Pareto front for 2-objective problems
+4.  **Access**: Access the latest Pareto front through `ga_instance.pareto_fronts` attribute
+
+### Example Usage
+```python
+import pygad
+import numpy as np
+
+def multi_objective_fitness(ga_instance, solution, solution_idx):
+    # Two objectives: maximize sum, minimize L2 norm
+    return [np.sum(solution), -np.sqrt(np.sum(solution**2))]
+
+# Create GA instance for multi-objective optimization
+ga_instance = pygad.GA(
+    num_generations=20,
+    num_parents_mating=5,
+    fitness_func=multi_objective_fitness,
+    sol_per_pop=20,
+    num_genes=5,
+    parent_selection_type="nsga2"
+)
+
+# Run the GA
+ga_instance.run()
+
+# Access the calculated Pareto fronts
+print(f"Number of Pareto fronts: {len(ga_instance.pareto_fronts)}")
+print(f"Best Pareto front size: {len(ga_instance.pareto_fronts[0])}")
+
+# Plot the Pareto front curve (for 2 objectives)
+ga_instance.plot_pareto_front_curve()
+```
+
 # Contact Us
 
 * E-mail: ahmed.f.gad@gmail.com

diff --git a/pygad/__init__.pyc b/pygad/__init__.pyc
diff --git a/pygad/__pycache__/__init__.cpython-313.pyc b/pygad/__pycache__/__init__.cpython-313.pyc
diff --git a/pygad/__pycache__/pygad.cpython-313.pyc b/pygad/__pycache__/pygad.cpython-313.pyc
diff --git a/pygad/helper/__pycache__/__init__.cpython-313.pyc b/pygad/helper/__pycache__/__init__.cpython-313.pyc
diff --git a/pygad/helper/__pycache__/misc.cpython-313.pyc b/pygad/helper/__pycache__/misc.cpython-313.pyc
diff --git a/pygad/helper/__pycache__/unique.cpython-313.pyc b/pygad/helper/__pycache__/unique.cpython-313.pyc
diff --git a/pygad/pygad.py b/pygad/pygad.py
diff --git a/pygad/utils/__pycache__/__init__.cpython-313.pyc b/pygad/utils/__pycache__/__init__.cpython-313.pyc
diff --git a/pygad/utils/__pycache__/crossover.cpython-313.pyc b/pygad/utils/__pycache__/crossover.cpython-313.pyc
diff --git a/pygad/utils/__pycache__/mutation.cpython-313.pyc b/pygad/utils/__pycache__/mutation.cpython-313.pyc
diff --git a/pygad/utils/__pycache__/nsga2.cpython-313.pyc b/pygad/utils/__pycache__/nsga2.cpython-313.pyc
diff --git a/pygad/utils/__pycache__/parent_selection.cpython-313.pyc b/pygad/utils/__pycache__/parent_selection.cpython-313.pyc
diff --git a/pygad/visualize/__pycache__/__init__.cpython-313.pyc b/pygad/visualize/__pycache__/__init__.cpython-313.pyc
diff --git a/pygad/visualize/__pycache__/plot.cpython-313.pyc b/pygad/visualize/__pycache__/plot.cpython-313.pyc
diff --git a/test_multi_objective.py b/test_multi_objective.py
@@ -0,0 +1,38 @@
+# -*- coding: utf-8 -*-
+import pygad
+import numpy
+import time
+
+def fitness_func(ga_instance, solution, solution_idx):
+    # Simple test fitness function
+    fitness_score = numpy.sum(solution * numpy.array([0.2, 0.3, 0.3, 0.2]))
+    # Simulate time cost
+    time_cost = numpy.sum(solution) * 0.1
+    # Simulate diversity score
+    diversity_score = numpy.random.uniform(0, 1)
+
+    # Return a composite fitness score for compatibility with existing GA implementation
+    # You can adjust the weights based on your priorities
+    composite_score = fitness_score + diversity_score - time_cost
+    return composite_score
+
+ga_instance = pygad.GA(num_generations=30,
+                       num_parents_mating=4,
+                       fitness_func=fitness_func,
+                       sol_per_pop=8,
+                       num_genes=4,
+                       parent_selection_type="sss",
+                       crossover_type="single_point",
+                       mutation_type="random",
+                       mutation_percent_genes=10)
+
+ga_instance.run()
+
+# 打印最终结果
+solution, solution_fitness, solution_idx = ga_instance.best_solution()
+print("\nFinal solution:", solution)
+print("\nFinal solution fitness:", solution_fitness)
+
+# 输出最后一代的Pareto前沿
+final_pareto = ga_instance.calculate_pareto_front(ga_instance.last_generation_fitness)
+print("\nFinal Pareto front size: {}".format(len(final_pareto)))
diff --git a/tests/__pycache__/test_adaptive_mutation.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_adaptive_mutation.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_adaptive_mutation.cpython-313.pyc b/tests/__pycache__/test_adaptive_mutation.cpython-313.pyc
diff --git a/tests/__pycache__/test_allow_duplicate_genes.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_allow_duplicate_genes.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_crossover_mutation.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_crossover_mutation.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_environment_switch.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_environment_switch.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_gene_constraint.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_gene_constraint.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_gene_space.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_gene_space.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_gene_space_allow_duplicate_genes.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_gene_space_allow_duplicate_genes.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_gene_type.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_gene_type.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_lifecycle_callbacks_calls.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_lifecycle_callbacks_calls.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_number_fitness_function_calls.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_number_fitness_function_calls.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_save_solutions.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_save_solutions.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_stop_criteria.cpython-313-pytest-9.0.2.pyc b/tests/__pycache__/test_stop_criteria.cpython-313-pytest-9.0.2.pyc
diff --git a/tests/__pycache__/test_stop_criteria.cpython-313.pyc b/tests/__pycache__/test_stop_criteria.cpython-313.pyc
diff --git a/tests/test_environment_switch.py b/tests/test_environment_switch.py
@@ -0,0 +1,251 @@
+"""
+This file contains tests for the environment switching feature in PyGAD.
+
+The environment switching feature allows PyGAD to automatically adjust genetic algorithm parameters
+based on different predefined environments:
+- ENV_STABLE: Default environment with standard parameters
+- ENV_FAST: Optimized for faster convergence
+- ENV_DIVERSE: Optimized for maintaining population diversity
+
+The tests verify:
+1. The default environment switching pattern (cycles every 10 generations)
+2. Correct parameter adjustments for each environment
+3. Proper handling of adaptive mutation parameters
+4. Multi-objective optimization with environment switching
+"""
+import pygad
+import numpy as np
+
+def test_environment_switch_pattern():
+    """
+    Test that the default environment switching pattern cycles correctly
+    every 10 generations: STABLE → FAST → DIVERSE → STABLE...
+    """
+    def fitness_func(ga_instance, solution, solution_idx):
+        return np.sum(solution)
+
+    ga_instance = pygad.GA(num_generations=1,
+                          num_parents_mating=2,
+                          fitness_func=fitness_func,
+                          sol_per_pop=5,
+                          num_genes=3)
+
+    # 测试环境切换模式
+    expected_pattern = [pygad.GA.ENV_STABLE,  # generation 0
+                        pygad.GA.ENV_STABLE,  # generations 1-9
+                        pygad.GA.ENV_FAST,    # generations 10-19
+                        pygad.GA.ENV_DIVERSE, # generations 20-29
+                        pygad.GA.ENV_STABLE,  # generations 30-39
+                        pygad.GA.ENV_FAST]    # generations 40-49
+
+    test_generations = [0, 5, 10, 20, 30, 40]
+
+    for gen, expected_env in zip(test_generations, expected_pattern):
+        env = ga_instance.get_environment_state(gen)
+        assert env == expected_env, f"Generation {gen}: expected {expected_env}, got {env}"
+
+def test_environment_parameters_adjustment():
+    """
+    Test that genetic algorithm parameters are adjusted correctly
+    when switching between different environments:
+    - FAST: Increase crossover probability by 20%, decrease mutation by 50%
+    - DIVERSE: Decrease crossover probability by 30%, increase mutation by 100%
+    - STABLE: Restore original parameters
+    """
+    def fitness_func(ga_instance, solution, solution_idx):
+        return np.sum(solution)
+
+    # 测试FAST环境
+    ga_instance = pygad.GA(num_generations=10,
+                          num_parents_mating=2,
+                          fitness_func=fitness_func,
+                          sol_per_pop=5,
+                          num_genes=3,
+                          crossover_probability=0.8,
+                          mutation_probability=0.1)
+
+    # 保存原始参数
+    original_crossover_rate = ga_instance.crossover_probability
+    original_mutation_rate = ga_instance.mutation_probability
+    ga_instance.original_selection_type = ga_instance.parent_selection_type
+
+    # 模拟FAST环境
+    ga_instance.current_environment = pygad.GA.ENV_FAST
+
+    # 应用FAST环境参数调整，对应run()方法中的内嵌逻辑
+    if isinstance(original_crossover_rate, (int, float)):
+        ga_instance.crossover_probability = original_crossover_rate * 1.2
+    else:
+        ga_instance.crossover_probability = original_crossover_rate
+
+    if isinstance(original_mutation_rate, (int, float)):
+        ga_instance.mutation_probability = original_mutation_rate * 0.5
+    else:
+        ga_instance.mutation_probability = original_mutation_rate
+
+    ga_instance.parent_selection_type = ga_instance.original_selection_type if ga_instance.original_selection_type not in ['nsga2', 'tournament_nsga2'] else 'tournament'
+
+    # 检查参数调整是否正确
+    assert ga_instance.crossover_probability == original_crossover_rate * 1.2, f"FAST environment crossover rate adjustment failed"
+    assert ga_instance.mutation_probability == original_mutation_rate * 0.5, f"FAST environment mutation rate adjustment failed"
+    assert ga_instance.parent_selection_type == "sss", f"FAST environment selection type adjustment failed"
+
+    # 模拟DIVERSE环境
+    ga_instance.current_environment = pygad.GA.ENV_DIVERSE
+
+    # 应用DIVERSE环境参数调整，对应run()方法中的内嵌逻辑
+    if isinstance(original_crossover_rate, (int, float)):
+        ga_instance.crossover_probability = original_crossover_rate * 0.7
+    else:
+        ga_instance.crossover_probability = original_crossover_rate
+
+    if isinstance(original_mutation_rate, (int, float)):
+        ga_instance.mutation_probability = original_mutation_rate * 2
+    else:
+        ga_instance.mutation_probability = original_mutation_rate
+
+    ga_instance.parent_selection_type = 'tournament_nsga2'
+
+    # 检查参数调整是否正确
+    assert ga_instance.crossover_probability == original_crossover_rate * 0.7, f"DIVERSE environment crossover rate adjustment failed"
+    assert ga_instance.mutation_probability == original_mutation_rate * 2, f"DIVERSE environment mutation rate adjustment failed"
+    assert ga_instance.parent_selection_type == "tournament_nsga2", f"DIVERSE environment selection type adjustment failed"
+
+    # 模拟回到STABLE环境
+    ga_instance.current_environment = pygad.GA.ENV_STABLE
+
+    # 恢复原始参数，对应run()方法中的内嵌逻辑
+    ga_instance.crossover_probability = original_crossover_rate
+    ga_instance.mutation_probability = original_mutation_rate
+    ga_instance.parent_selection_type = ga_instance.original_selection_type
+
+    # 检查参数是否恢复原值
+    assert ga_instance.crossover_probability == original_crossover_rate, f"STABLE environment crossover rate should remain unchanged"
+    assert ga_instance.mutation_probability == original_mutation_rate, f"STABLE environment mutation rate should remain unchanged"
+    assert ga_instance.parent_selection_type == "sss", f"STABLE environment selection type should remain unchanged"
+
+def test_sequence_parameters_in_adaptive_mutation():
+    """
+    Test that sequence parameters (like adaptive mutation rates) remain unchanged
+    when switching environments. Adaptive mutation uses a sequence of rates
+    that should not be overwritten by environment adjustments.
+    """
+    def fitness_func(ga_instance, solution, solution_idx):
+        return np.sum(solution)
+
+    ga_instance = pygad.GA(num_generations=10,
+                          num_parents_mating=2,
+                          fitness_func=fitness_func,
+                          sol_per_pop=5,
+                          num_genes=3,
+                          mutation_probability=[0.2, 0.05],
+                          mutation_type="adaptive")
+
+    original_mutation_rate = ga_instance.mutation_probability.copy()
+
+    # 在FAST环境下测试，自适应突变参数应该保持不变
+    ga_instance.current_environment = pygad.GA.ENV_FAST
+
+    # 应用参数调整，对应run()方法中的内嵌逻辑
+    original_mutation_rate = ga_instance.mutation_probability
+    if isinstance(original_mutation_rate, (int, float)):
+        ga_instance.mutation_probability = original_mutation_rate * 0.5
+    else:
+        ga_instance.mutation_probability = original_mutation_rate
+
+    # 验证自适应突变参数未被修改
+    assert ga_instance.mutation_probability == original_mutation_rate, "Adaptive sequence mutation rate should remain unchanged in FAST environment"
+
+    # 在DIVERSE环境下测试
+    ga_instance.current_environment = pygad.GA.ENV_DIVERSE
+
+    # 应用参数调整，对应run()方法中的内嵌逻辑
+    if isinstance(original_mutation_rate, (int, float)):
+        ga_instance.mutation_probability = original_mutation_rate * 2
+    else:
+        ga_instance.mutation_probability = original_mutation_rate
+
+    # 验证自适应突变参数仍未被修改
+    assert ga_instance.mutation_probability == original_mutation_rate, "Adaptive sequence mutation rate should remain unchanged in DIVERSE environment"
+
+def test_multi_objective_environment_switch():
+    """
+    Test environment switching with multi-objective optimization.
+    Verifies that the environment switching works correctly with
+    non-single objective fitness functions and proper Pareto front handling.
+    """
+    def fitness_func(ga_instance, solution, solution_idx):
+        # Multi-objective fitness: maximize sum, minimize L2 norm
+        return [np.sum(solution), -np.sqrt(np.sum(solution**2))]
+
+    ga_instance = pygad.GA(num_generations=30,
+                          num_parents_mating=2,
+                          fitness_func=fitness_func,
+                          sol_per_pop=10,
+                          num_genes=3)
+
+    # 运行遗传算法，验证环境切换正常工作且不报错
+    ga_instance.run()
+
+    # 验证Pareto前沿被正确计算
+    assert hasattr(ga_instance, 'pareto_fronts'), "Pareto fronts attribute should exist"
+    assert len(ga_instance.pareto_fronts) > 0, "Pareto fronts should not be empty"
+
+def test_three_dimensional_multi_objective():
+    """
+    Test multi-objective optimization with three-dimensional fitness vectors
+    to verify proper Pareto front calculation for higher-dimensional problems.
+    """
+    def fitness_func(ga_instance, solution, solution_idx):
+        # Three objectives:
+        # 1. Maximize sum of solution
+        # 2. Minimize L2 norm
+        # 3. Maximize product of non-zero elements
+        sum_obj = np.sum(solution)
+        norm_obj = -np.sqrt(np.sum(solution**2))
+        product_obj = np.prod(solution[solution != 0]) if np.any(solution != 0) else 0
+        return [sum_obj, norm_obj, product_obj]
+
+    ga_instance = pygad.GA(
+        num_generations=10,
+        num_parents_mating=5,
+        fitness_func=fitness_func,
+        sol_per_pop=20,
+        num_genes=5,
+        parent_selection_type="nsga2",
+        crossover_probability=0.8,
+        mutation_probability=0.1
+    )
+
+    # Run for a few generations
+    ga_instance.run()
+
+    # Verify Pareto front was calculated correctly for 3 objectives
+    assert hasattr(ga_instance, 'pareto_fronts'), "Pareto fronts attribute should exist"
+    assert len(ga_instance.pareto_fronts) > 0, "Pareto fronts should not be empty"
+
+    # Check that each solution in the Pareto front has 3 objectives
+    for front in ga_instance.pareto_fronts:
+        for solution in front:
+            if len(solution) >= 2:
+                # solution array contains [solution_idx, fitness_values]
+                fitness = solution[1]
+                assert len(fitness) == 3, "3-objective fitness expected"
+
+if __name__ == "__main__":
+    test_environment_switch_pattern()
+    print("✓ test_environment_switch_pattern passed")
+
+    test_environment_parameters_adjustment()
+    print("✓ test_environment_parameters_adjustment passed")
+
+    test_sequence_parameters_in_adaptive_mutation()
+    print("✓ test_sequence_parameters_in_adaptive_mutation passed")
+
+    test_multi_objective_environment_switch()
+    print("✓ test_multi_objective_environment_switch passed")
+
+    test_three_dimensional_multi_objective()
+    print("✓ test_three_dimensional_multi_objective passed")
+
+    print("\nAll tests passed!")