使用 scipy.optimize 通过多个变量和边界最小化函数

Question

我要优化这个功能：

def cost_func(adg_guess,wave_guess,wave_opt,C0,C1,S):
    ka=0
    for wave in wave_guesse:
        ka=ka+(adg_guess[find(wave,wave_guess)]-C0*np.exp(-S*(wave-440))-C1)^2
    return ka

from scipy.optimize import minimize
from scipy.optimize import Bounds
def anw_deco(anw_no763,a0,a1,wave_full,wave_opt):
    adg380_init,aph380_init=anw_de_gues(anw_no763)
    aph_gu=aph_guess(wave_full,a0,a1,aph380_init)
    adg_gu=adg_guess(anw_no763,aph_gu)
    cost_func_loca=lambda C0,C1,S:cost_func(C0=C0,
                                            C1=C1,
                                            S=S,
                                            adg_guess=adg_gu,
                                             wave_guess=wave_full,
                                             wave_opt=wave_opt)
    bonds=Bounds([0,0,0],[np.inf,np.inf,0.03])
    res,_ = minimize(cost_func_loca(C0,C1,S),x0=np.array([1,1,0.01]),bounds=bonds, tol=1e-6, options={'maxiter': 1e3},)
    C0=res[1]
    C1=res[2]
    S=res[3]
    dlta=dta(adg_guess,S,C0,C1,wave_full)
    if np.mean(dlta)<0.01:
        adg_fit=C0*np.exp(-S*(wave_full-440))+C1
        return anw_no763,adg_fit,S
    else:
        anw_deco(anw_no763-dlta)

其他函数用于猜测结果并计算停止条件。

但是当我想像这样传递代码时

anw_deco(anw_no763=anwtest,
        a0=a0_aphfit,
        a1=a1_aphfit,
        wave_full=SGLI_waveno763,
        wave_opt=wave_for_opt)

它建议我

TypeError: <lambda>() missing 2 required positional arguments: 'C1' and 'S'

有什么问题？

Answer 1

minimize 的函数必须接收一维 numpy 数组。尝试解压缩函数中的 C0,C1,S 变量：

def cost_func_loca(x) =
    C0,C1,S=x
    return cost_func(C0=C0,
              C1=C1,
              S=S,
              adg_guess=adg_gu,
              wave_guess=wave_full,
              wave_opt=wave_opt)

您还必须将函数对象传递给minimize（即不带参数），如下所示：

res,_ = minimize(cost_func_loca,x0=np.array([1,1,0.01]),bounds=bonds, tol=1e-6, options={'maxiter': 1e3},)