人工蜂コロニーアルゴリズムで関数の最小値探索を行う

[コードを表示]

Python

import numpy as np
import matplotlib.pyplot as plt
import math

# 関数の設定
# xはndarray

# sphere
def func(x):
    return np.sum(x**2)

# rastrigin
# def func(x):
#     s = 0
#     for i in range(len(x)):
#         s += x[i]**2 - 10*math.cos(2*math.pi*x[i])
#     return 10*d + s


# 初期設定
N = 100 # 収穫蜂と追従蜂の個体数
d = 2 # 次元
TC = np.zeros(N) #更新カウント
lim = 30
xmax = 5
xmin = -5
G = 100 # 繰り返す回数

x = np.zeros((N,d))
for i in range(N):
    x[i] = (xmax-xmin)*np.random.rand(d) + xmin

def fit(x):
    z = func(x)
    if z > 0:
        return 1/(1+z)
    else: return 1+abs(z)

# ルーレット選択用関数
def roulette_choice(w):
    tot = []
    acc = 0
    for e in w:
        acc += e
        tot.append(acc)

    r = np.random.random() * acc
    for i, e in enumerate(tot):
        if r <= e:
            return i

x_best = [0,0] #x_bestの初期化
best = 100

# 繰り返し
best_value = []
ims = []
for g in range(G):
    best_value.append(best)

    z = np.zeros(N)
    for i in range(N):
        z[i] = func(x[i])

    # employee bee step
    for i in range(N):
        v = x.copy()
        k = i
        while k == i:
            k = np.random.randint(N)
        j = np.random.randint(d)
        r = np.random.rand()*2 - 1 # -1から1までの一様乱数
        v[i,j] = x[i,j] + r * (x[i,j] - x[k,j])

        if fit(x[i]) < fit(v[i]):
            x[i] = v[i]
            TC[i] = 0
        else: TC[i] += 1

    # onlooker bee step
    w = []
    for i in range(N):
        w.append(fit(x[i]))
    
    for l in range(N):
        i = roulette_choice(w)
        j = np.random.randint(d)
        r = np.random.rand()*2 - 1 # -1から1までの一様乱数
        v[i,j] = x[i,j] + r * (x[i,j] - x[k,j])
        if fit(x[i]) < fit(v[i]):
            x[i] = v[i]
            TC[i] = 0
        else: TC[i] += 1


    # scout bee step
    for i in range(N):
        if TC[i] > lim:
            for j in range(d):
                x[i,j] = np.random.rand()*(xmax-xmin) + xmin
            TC[i] = 0

    # 最良個体の発見
    for i in range(N):
        if best > func(x[i]):
            x_best = x[i].copy()
            best = func(x[i])

# 結果
print("発見した最小値：{}\nそのときのx：{}".format(func(x_best),x_best))
plt.plot(range(len(best_value)),best_value)
plt.yscale('log')
plt.title("Sphere")
plt.xlabel("試行回数")
plt.ylabel("発見した最小値")
plt.show()