我有一個數據集,我想通過最小平方誤差方法找到混合高斯模型。Scipy&Optimize:最小化示例,如何添加約束?
的代碼是這樣的:
from sklearn.neighbors import KernelDensity
kde = KernelDensity().fit(sample)
def gaussian_2d(x,y,meanx,meany,sigx,sigy,rho):
# rho <= 1
part1 = 1/(2*np.pi*sigx*sigy*sqrt(1-0.5**2))
part2 = -1/2*(1-rho**2)
part3 = (((x-meanx)/sigx)**2-2*rho*(x-meanx)*(y-meany)/(sigx*sigy)+((y-meany)/sigy)**2)
return part1*exp(part2*part3)
def square_error(f1,f2, u1,v1,sigu1,sigv1,rho1, u2,v2,sigu2,sigv2,rho2, u3,v3,sigu3,sigv3,rho3):
# 1. Generate Mixed Gaussian Model
def gaussian1(x,y):
return gaussian_2d(x,y,u1,v1,sigu1,sigv1,rho1)
def gaussian2(x,y):
return gaussian_2d(x,y,u2,v2,sigu2,sigv2,rho2)
def gaussian3(x,y):
return gaussian_2d(x,y,u3,v3,sigu3,sigv3,rho3)
mixed_model = f1*gaussian1(x,y)+f2*gaussian2(x,y)+(1-f1-f2)*gaussian3(x,y)
# 2. Calculate the sum of square error
sum_error = 0
for point in sample:
error = (exp(mixed_model(point)) - exp(kde.score(point)))**2
sum_error += error
return sum_error
# How can I add constraints:
# f1+f2 <= 1
# rho1,2,3 <= 1
result = sp.optimize.minimize(square_error)
但我不知道如何在minimize
方法添加約束。 http://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html#scipy.optimize.minimize中的示例很難理解。
UPDATE: 這是我結束了,
result = sp.optimize.minimize(
square_error,
x0 = [0.2,0.5,
1,1,1,1,0.3,
1,1,1,1,0.3,
1,1,1,1,0.3,],
bounds = [(0., 1.),(0., 1.),
(None, None),(None, None),(0., None),(0., None),(0., 1.),
(None, None),(None, None),(0., None),(0., None),(0., 1.),
(None, None),(None, None),(0., None),(0., None),(0., 1.),])
但它給了我TypeError: square_error() takes exactly 17 arguments (1 given)
,有什麼問題呢?
可能重複的[scipy最小化約束](http://stackoverflow.com/questions/20075714/scipy-minimize-with-constraints) –