-
Notifications
You must be signed in to change notification settings - Fork 0
/
ABC.py
122 lines (106 loc) · 5.86 KB
/
ABC.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
import numpy as np
import random, math, copy,time
import matplotlib.pyplot as plt
from fun import fu
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
def GrieFunc(data): # 目标函数
y = fu(data[0],int(data[1]),int(data[2]),data[3])
return y
class ABSIndividual:
def __init__(self, bound):
self.score = 0.
self.invalidCount = 0 # 无效次数（成绩没有更新的累积次数）
self.bound=bound
self.chrom = [random.uniform(self.bound[0, 0], self.bound[1, 0]), random.randint(self.bound[0, 1], self.bound[1, 1]), random.randint(self.bound[0, 2], self.bound[1, 2]), random.uniform(self.bound[0, 3], self.bound[1, 3])] # 随机初始化
self.calculateFitness()
def calculateFitness(self):
self.score = GrieFunc(self.chrom) # 计算当前适应度
class ArtificialBeeSwarm:
def __init__(self, foodCount, onlookerCount, bound, maxIterCount, maxInvalidCount):
self.foodCount = foodCount # 蜜源个数，等同于雇佣蜂数目
self.onlookerCount = onlookerCount # 观察蜂个数
self.bound = bound # 各参数上下界
self.maxIterCount = maxIterCount # 迭代次数
self.maxInvalidCount = maxInvalidCount # 最大无效次数
self.foodList = [ABSIndividual(self.bound) for k in range(self.foodCount)] # 初始化各蜜源
self.foodScore = [d.score for d in self.foodList] # 各蜜源最佳成绩
self.bestFood = self.foodList[np.argmax(self.foodScore)] # 全局最佳蜜源
def updateFood(self, i): # 更新第i个蜜源
j = random.choice([d for d in range(self.foodCount) if d != i])
vi = copy.deepcopy(self.foodList[i])
vi.chrom[0] = self.foodList[i].chrom[0] + random.uniform(-1.0, 1.0) * (self.foodList[i].chrom[0] - self.foodList[j].chrom[0])
vi.chrom[1] = int(self.foodList[i].chrom[1] + random.uniform(-1.0, 1.0) * (self.foodList[i].chrom[1] - self.foodList[j].chrom[1]))
vi.chrom[2] = int(self.foodList[i].chrom[2] + random.uniform(-1.0, 1.0) * (self.foodList[i].chrom[2] - self.foodList[j].chrom[2]))
vi.chrom[3] = self.foodList[i].chrom[3] + random.uniform(-1.0, 1.0) * (self.foodList[i].chrom[3] - self.foodList[j].chrom[3])
vi.chrom[0] = np.clip(vi.chrom[0], self.bound[0, 0], self.bound[1, 0]) # 参数不能越界
vi.chrom[1] = int(np.clip(vi.chrom[1], self.bound[0, 1], self.bound[1, 1])) # 参数不能越界
vi.chrom[2] = int(np.clip(vi.chrom[2], self.bound[0, 2], self.bound[1, 2])) # 参数不能越界
vi.chrom[3] = np.clip(vi.chrom[3], self.bound[0,3], self.bound[1, 3]) # 参数不能越界
vi.calculateFitness()
if vi.score > self.foodList[i].score: # 如果成绩比当前蜜源好
self.foodList[i] = vi
if vi.score > self.foodScore[i]: # 如果成绩比历史成绩好（如重新初始化，当前成绩可能低于历史成绩）
self.foodScore[i] = vi.score
if vi.score > self.bestFood.score: # 如果成绩全局最优
self.bestFood = vi
self.foodList[i].invalidCount = 0
else:
self.foodList[i].invalidCount += 1
def employedBeePhase(self):
for i in range(0, self.foodCount): # 各蜜源依次更新
self.updateFood(i)
def onlookerBeePhase(self):
foodScore = [d.score for d in self.foodList]
maxScore = np.max(foodScore)
accuFitness = [(0.9 * d / maxScore + 0.1, k) for k, d in enumerate(foodScore)] # 得到各蜜源的 相对分数和索引号
for k in range(0, self.onlookerCount):
i = random.choice([d[1] for d in accuFitness if d[0] >= random.random()]) # 随机从相对分数大于随机门限的蜜源中选择跟随
self.updateFood(i)
def scoutBeePhase(self):
for i in range(0, self.foodCount):
if self.foodList[i].invalidCount > self.maxInvalidCount: # 如果该蜜源没有更新的次数超过指定门限，则重新初始化
self.foodList[i] = ABSIndividual(self.bound)
self.foodScore[i] = max(self.foodScore[i], self.foodList[i].score)
# print(self.foodScore)
def solve(self):
trace = []
trace.append((self.bestFood.score, np.mean(self.foodScore)))
start = time.time()
for k in range(self.maxIterCount):
self.employedBeePhase()
self.onlookerBeePhase()
self.scoutBeePhase()
trace.append((self.bestFood.score, np.mean(self.foodScore)))
end = time.time()
print(str(end - start))
print(self.bestFood.score,self.bestFood.chrom)
self.printResult(np.array(trace))
def printResult(self, trace):
x = np.arange(0, trace.shape[0])
print(trace[:, 0])
plt.plot(x, [d for d in trace[:, 0]], 'b', label='optimal value')
plt.plot(x, [d for d in trace[:, 1]], 'g', label='average value')
plt.xlabel("Iteration")
plt.ylabel("function value(log)")
plt.yscale('log')
plt.title("Artificial Bee Swarm algorithm for function optimization")
plt.legend()
plt.show()
if __name__ == "__main__":
random.seed()
max_lr = 0.01
min_lr = 0.00001
max_batch_size = 256
min_batch_size = 4
max_CLIENT_EPOCHS = 10
min_CLIENT_EPOCHS = 1
max_momentum= 1
min_momentum=0.00001
foodCount = 5
onlookerCount = foodCount
maxIterCount = 5
maxInvalidCount = 3
bound = np.array([[min_lr,min_batch_size,min_CLIENT_EPOCHS,min_momentum], [max_lr,max_batch_size,max_CLIENT_EPOCHS,max_momentum]]) #默认np.tile(参数1是要扩充的实体, 参数2是沿着某个方向进行扩充)
abc = ArtificialBeeSwarm(foodCount, onlookerCount, bound, maxIterCount, maxInvalidCount)
abc.solve()