hooke jeeves算法（模式搜索法）

# -*- coding: utf-8 -*-
"""
Spyder Editor

This is a temporary script file.
"""

import numpy as np
import matplotlib.pyplot as plt

# -----------define HJ agolrathiem para
delta = 0.5 # step size
alfa = 1    # speed rate
beta = 0.5  # decrease rate
epsilon = 0.002 # min stepsize
iniPO = np.array([6,1])     # initial point
G = 10  # 探测次数

# -----------define function
# min f(x):=(1−x1​)^2+5(x2​−x1^​2)^2
def fun(A):
    ff = (1 - A[0]) ** 2 + 5 * (A[1] - A[0] ** 2) ** 2
    return ff

iniPO_ff = np.array([fun(iniPO)])
X = iniPO
X_ff = iniPO_ff
Y = X
Y_ff =X_ff
x = iniPO
y = iniPO

# -----------explore step
dim = iniPO.shape[0] # dimation
e = [1]*dim
E = np.diag(e)
k=0
while delta > epsilon:    
    for i in range(0,dim):
       y_new_plus = y + delta * E[i,:]
       ff = fun(y)
       ff_new = np.array([fun(y_new_plus)])
       temp = Y
       temp_ff = Y_ff
       Y = np.vstack((temp,y_new_plus))
       Y_ff = np.vstack((temp_ff,ff_new))
       plt.plot([Y_ff.shape[0]-1,Y_ff.shape[0]], Y_ff[-2:,:], c='b', marker='*')
       #plt.show()
       if ff_new < ff:
           y = y_new_plus   
       else:
           y_new_minus = y - delta * E[i,:]
           ff = fun(y)
           ff_new = np.array([fun(y_new_minus)])
           temp = Y
           temp_ff = Y_ff
           Y = np.vstack((temp,y_new_minus))
           Y_ff = np.vstack((temp_ff,ff_new))
           plt.plot([Y_ff.shape[0]-1,Y_ff.shape[0]], Y_ff[-2:,:], c='b', marker='*')
           #plt.show()
           if ff_new < ff:
               y = y_new_minus                 
    # get all points
    Y_Data = np.hstack((Y,Y_ff))
    # get new base point
    x =  y
    # get all base points
    temp_x = X
    X = np.vstack((temp_x,x))
    temp_x_ff = X_ff
    X_ff = np.vstack((temp_x_ff,fun(x)))
    X_Data = np.hstack((X,X_ff))
 
# -----------Move step  
    
    if ff_new < ff:
        print()
        y = X[-1,:] + alfa* (X[-1,:]-X[-2,:])
        ff = fun(y)
        temp = Y
        Y = np.vstack((temp,y))
        temp_ff = Y_ff
        Y_ff = np.vstack((Y_ff,ff))
        plt.plot([Y_ff.shape[0]-1,Y_ff.shape[0]], Y_ff[-2:,:], c='b', marker='*')
        Y_Data = np.hstack((Y,Y_ff))  
    else:
        #temp = Y
        #Y = np.vstack((temp,y))
        #temp_ff = Y_ff
        #Y_ff = np.vstack((Y_ff,ff))
        #plt.plot([Y_ff.shape[0]-1,Y_ff.shape[0]], Y_ff[-2:,:], c='b', marker='*')
        #Y_Data = np.hstack((Y,Y_ff))
        delta = beta * delta
    print('k',k)
    print('')
#-----------wheather stop
    if k >= G:     # 当k大于等于G跳出循环
        break
    k = k+1
    
print('--------')
print('Y_Data',Y_Data)
print('X_Data',X_Data)
print('delta',delta)
print('--------')