模拟退火算法（Simulated Annealing，SA）的全面讲解及python实现

import numpy as np
import as plt
import random


class SA(object):
    def __init__(self, interval, tab='min', T_max=1000, T_min=1, iterMax=1000, rate=0.95):
        self.interval = interval  # 状态空间，即解的搜索范围
        self.T_max = T_max  # 温度上限
        self.T_min = T_min  # 温度下限
        self.iterMax = iterMax  # 定温内部迭代次数
        self.rate = rate  # 退火降温速度
        self.x_seed = (interval[0], interval[1])
        self.tab = ()  # 求解最大值还是最小值的标签:‘min','max'
        self.solve()
        self.display()

    def solve(self):
        temp = 'deal_' + self.tab
        if hasattr(self, temp):
            deal = getattr(self, temp)  # 采用反射方法提取对应的函数
        else:
            exit(">>>tab标签传参有误：'min'|'max'<<<")
        x1 = self.x_seed
        T = self.T_max
        while T >= self.T_min:
            for i in range(self.iterMax):
                f1 = self.func(x1)
                delta_x = ()*2-1
                if self.interval[0] <= x1+delta_x <= self.interval[1]:
                    x2 = x1 + delta_x
                else:
                    x2 = x1 - delta_x
                f2 = self.func(x2)
                delta_f = f2 - f1
                x1 = deal(x1, x2, delta_f, T)
            T *= self.rate
        self.x_solu = x1
        self.f_solu = self.func(x1)

    def func(self, x):
        value = (x**2)*(x**2-5*x)
        return value

    def p_min(self, delta, T):
        probability = (-delta/T)
        return probability

    def p_max(self, delta, T):
        probability = (delta/T)
        return probability

    def deal_min(self, x1, x2, delta, T):
        if delta < 0:
            return x2
        else:
            p = self.p_min(delta,T)
            if p > ():
                return x2
            else:
                return x1

    def deal_max(self, x1, x2, delta, T):
        if delta > 0:
            return x2
        else:
            p = self.p_max(delta, T)
            if p > ():
                return x2
            else:
                return x1

    def display(self):
        print(f'seed: {self.x_seed}\nseed_funtion value: {self.func(self.x_seed)}\nsolution: {self.x_solu}\nfunction value: {self.f_solu}')
        (figsize=(6, 4))
        x = (self.interval[0], self.interval[1])
        y = self.func(x)
        (x,y,'g-',label='function')
        (self.x_seed,self.func(self.x_seed),'bo',label='seed')
        (self.x_solu,self.func(self.x_solu),'r*',label='solution')
        (f'solution={self.x_solu}')
        ('x')
        ('y')
        ()
        ()
        ()

if __name__ == '__main__':
    SA([-5,5],'max')