模拟退火的粒子群算法python代码

模拟退火粒子群优化（Simulated Annealing Particle Swarm Optimization, SAPSO）是一种结合了模拟退火算法和粒子群优化（PSO）策略的搜索方法。它通过模拟金属冷却过程中的相变过程，允许在局部最优解附近探索全局最优解。以下是简单的Python实现框架：

import numpy as np

class SimulatedAnnealingPSO:
    def __init__(self, n_particles, dim, bounds, max_iter, t_initial, t_min, alpha):
        self.n_particles = n_particles
        self.dim = dim
        self.bounds = bounds
        self.max_iter = max_iter
        self.t_initial = t_initial
        self.t_min = t_min
        self.alpha = alpha
        self.particles = np.random.uniform(bounds[0], bounds[1], (n_particles, dim))
        self.velocities = np.zeros_like(self.particles)
        self.best_positions = self.particles.copy()
        self.global_best_position = self.particles[np.argmax(np.sum(self.particles**2, axis=1))]

    def cooling_schedule(self, iteration):
        return self.t_initial * np.exp(-iteration / self.max_iter)

    def move_particles(self, particles, velocities, best_positions, global_best_position, temperature):
        for i in range(len(particles)):
            r1, r2 = np.random.rand(2)
            velocities[i] = self.inertia_term(velocities[i], r1) + \
                            c1 * r2 * (best_positions[i] - particles[i]) + \
                            c2 * r2 * (global_best_position - particles[i])
            if np.random.rand() < np.exp(-(np.sum((velocities[i] + particles[i]) ** 2) - np.sum(particles[i]**2)) / (2 * temperature)):
                particles[i] += velocities[i]
                if not self.is_valid(particles[i]):
                    particles[i] = self.resample(particles[i])

    def inertia_term(self, velocity, r):
        return velocity * np.where(r < 0.5, 2*r, 2*(1-r))

    # 辅助函数：检查新位置是否有效，并进行随机重采样
    def is_valid(self, position):
        return all(position >= self.bounds[0]) and all(position <= self.bounds[1])

    def resample(self, position):
        valid_pos = np.clip(position, self.bounds[0], self.bounds[1])
        return valid_pos

    def optimize(self):
        for iteration in range(self.max_iter):
            temperature = self.cooling_schedule(iteration)
            self.move_particles(self.particles, self.velocities, self.best_positions, self.global_best_position, temperature)
            self.update_best_positions()
        return self.global_best_position

# 使用示例：
n_particles = 50
dim = 10
bounds = (-5, 5)
max_iter = 1000
t_initial = 1
t_min = 0.01
alpha = 0.9

algorithm = SimulatedAnnealingPSO(n_particles, dim, bounds, max_iter, t_initial, t_min, alpha)
result = algorithm.optimize()
print("全局最佳解:", result)

这个代码创建了一个`SimulatedAnnealingPSO`类，包含了粒子初始化、温度调整、移动粒子以及更新最佳位置等核心步骤。你可以根据需要调整参数进行实际优化。