|
import numpy as np |
|
import random |
|
import math |
|
import time |
|
from typing import Callable, List, Tuple |
|
import matplotlib.pyplot as plt |
|
|
|
|
|
class QuantumInspiredMultiObjectiveOptimizer: |
|
def __init__(self, |
|
objective_fns: List[Callable[[List[float]], float]], |
|
dimension: int, |
|
population_size: int = 100, |
|
iterations: int = 200, |
|
tunneling_prob: float = 0.2, |
|
entanglement_factor: float = 0.5, |
|
mutation_scale: float = 1.0, |
|
archive_size: int = 200): |
|
self.objective_fns = objective_fns |
|
self.dimension = dimension |
|
self.population_size = population_size |
|
self.iterations = iterations |
|
self.tunneling_prob = tunneling_prob |
|
self.entanglement_factor = entanglement_factor |
|
self.mutation_scale = mutation_scale |
|
self.archive_size = archive_size |
|
|
|
self.population = [self._random_solution() for _ in range(population_size)] |
|
self.pareto_front = [] |
|
self.archive = [] |
|
|
|
def _random_solution(self) -> List[float]: |
|
return [random.uniform(-10, 10) for _ in range(self.dimension)] |
|
|
|
def _tunnel(self, solution: List[float], scale: float) -> List[float]: |
|
return [x + np.random.normal(0, scale) * random.choice([-1, 1]) |
|
if random.random() < self.tunneling_prob else x |
|
for x in solution] |
|
|
|
def _entangle(self, solution1: List[float], solution2: List[float], factor: float) -> List[float]: |
|
return [(1 - factor) * x + factor * y for x, y in zip(solution1, solution2)] |
|
|
|
def _evaluate(self, solution: List[float]) -> List[float]: |
|
return [fn(solution) for fn in self.objective_fns] |
|
|
|
def _dominates(self, obj1: List[float], obj2: List[float]) -> bool: |
|
return all(o1 <= o2 for o1, o2 in zip(obj1, obj2)) and any(o1 < o2 for o1, o2 in zip(obj1, obj2)) |
|
|
|
def _pareto_selection(self, scored_population: List[Tuple[List[float], List[float]]]) -> List[Tuple[List[float], List[float]]]: |
|
pareto = [] |
|
for candidate in scored_population: |
|
if not any(self._dominates(other[1], candidate[1]) for other in scored_population if other != candidate): |
|
pareto.append(candidate) |
|
unique_pareto = [] |
|
seen = set() |
|
for sol, obj in pareto: |
|
key = tuple(round(x, 6) for x in sol) |
|
if key not in seen: |
|
unique_pareto.append((sol, obj)) |
|
seen.add(key) |
|
return unique_pareto |
|
|
|
def _update_archive(self, pareto: List[Tuple[List[float], List[float]]]): |
|
combined = self.archive + pareto |
|
combined = self._pareto_selection(combined) |
|
self.archive = sorted(combined, key=lambda x: tuple(x[1]))[:self.archive_size] |
|
|
|
def optimize(self) -> Tuple[List[Tuple[List[float], List[float]]], float]: |
|
start_time = time.time() |
|
for i in range(self.iterations): |
|
adaptive_tunnel = self.mutation_scale * (1 - i / self.iterations) |
|
adaptive_entangle = self.entanglement_factor * (1 - i / self.iterations) |
|
|
|
scored_population = [(sol, self._evaluate(sol)) for sol in self.population] |
|
pareto = self._pareto_selection(scored_population) |
|
self._update_archive(pareto) |
|
self.pareto_front = pareto |
|
|
|
new_population = [p[0] for p in pareto] |
|
while len(new_population) < self.population_size: |
|
parent1 = random.choice(pareto)[0] |
|
parent2 = random.choice(pareto)[0] |
|
if parent1 == parent2: |
|
parent2 = self._tunnel(parent2, adaptive_tunnel) |
|
child = self._entangle(parent1, parent2, adaptive_entangle) |
|
child = self._tunnel(child, adaptive_tunnel) |
|
new_population.append(child) |
|
|
|
self.population = new_population |
|
|
|
duration = time.time() - start_time |
|
return self.archive, duration |
|
|
|
|
|
def sphere(x: List[float]) -> float: |
|
return sum(xi ** 2 for xi in x) |
|
|
|
|
|
def rastrigin(x: List[float]) -> float: |
|
return 10 * len(x) + sum(xi ** 2 - 10 * math.cos(2 * math.pi * xi) for xi in x) |
|
|
|
|
|
if __name__ == '__main__': |
|
optimizer = QuantumInspiredMultiObjectiveOptimizer( |
|
objective_fns=[sphere, rastrigin], |
|
dimension=20, |
|
population_size=100, |
|
iterations=200, |
|
tunneling_prob=0.2, |
|
entanglement_factor=0.5, |
|
mutation_scale=1.0, |
|
archive_size=300 |
|
) |
|
|
|
pareto_front, duration = optimizer.optimize() |
|
print(f"Optimization completed in {duration:.2f} seconds") |
|
print(f"Pareto front size: {len(pareto_front)}") |
|
for sol, scores in pareto_front: |
|
print("Solution:", sol, "Objectives:", scores) |
|
|
|
if len(pareto_front[0][1]) == 2: |
|
x_vals = [obj[0] for _, obj in pareto_front] |
|
y_vals = [obj[1] for _, obj in pareto_front] |
|
plt.scatter(x_vals, y_vals, c='blue', label='Pareto Front') |
|
plt.xlabel('Objective 1') |
|
plt.ylabel('Objective 2') |
|
plt.title('Pareto Front Visualization') |
|
plt.legend() |
|
plt.grid(True) |
|
plt.show() |
|
|