def calReferObjV(self): # 设定目标数参考值（本问题目标函数参考值设定为理论最优值，即“真实帕累托前沿点”）
N = 1000 # 欲生成10000个全局帕累托最优解
# 参数a,b,c为求解方程得到，详见DTLZ7的参考文献
a = 0.2514118360889171
b = 0.6316265307000614
c = 0.8594008566447239
Vars, Sizes = ea.crtgp(self.M - 1, N) # 生成单位超空间内均匀的网格点集
middle = 0.5
left = Vars <= middle
right = Vars > middle
maxs_Left = np.max(Vars[left])
if maxs_Left > 0:
Vars[left] = Vars[left] / maxs_Left * a
Vars[right] = (Vars[right] - middle) / (np.max(Vars[right]) - middle) * (c - b) + b
P = np.hstack([Vars, (2 * self.M - np.sum(Vars * (1 + np.sin(3 * np.pi * Vars)), 1, keepdims=True))])
referenceObjV = P
return referenceObjV
def calBest(self): # 计算全局最优解
N = 1000 # 欲生成10000个全局帕累托最优解
# 参数a,b,c为求解方程得到，详见DTLZ7的参考文献
a = 0.2514118360889171
b = 0.6316265307000614
c = 0.8594008566447239
Vars, Sizes = ea.crtgp(self.M - 1, N) # 生成单位超空间内均匀的网格点集
middle = 0.5
left = Vars <= middle
right = Vars > middle
maxs_Left = np.max(Vars[left])
if maxs_Left > 0:
Vars[left] = Vars[left] / maxs_Left * a
Vars[right] = (Vars[right] - middle) / (np.max(Vars[right]) -
middle) * (c - b) + b
P = np.hstack([
Vars,
(2 * self.M -
np.sum(Vars * (1 + np.sin(3 * np.pi * Vars)), 1, keepdims=True))
])
globalBestObjV = P
return globalBestObjV