时间戳

首页 > 解决方案 > 每个数据的 Scipy.optimize.minimize

iteration - 每个数据的 Scipy.optimize.minimize

问题描述

我有 2 个模型的 1000 个预测结果和它们的真实值。对于每个数据，我必须在这个等式中找到 b0 和 b1： sqrt((w[i]**(b0+b1*MODEL1[i])*MODEL1[i]+(1-w[i])**(b0+b1*MODEL1[i])*MODEL2[i]-TRUE[i])**2)

在我的模型中，我从外部带来 w[i] 值。对于每个数据，我还有所有 ADB[i] (MODEL1)、RF[i] (MODEL2) 和 test_y[i] (TRUE)。下面的代码只能解决数据集中的一个数据。但是，我必须为 1000 个数据找到它，所以在优化结束时有 1000 b0 和 1000 b1。

def objective(b):
    b0 = b[0]
    b1 = b[1]
    return sqrt((w1[0]**(b0+b1*ADB[0])*ADB[0]+(1-w1[0])**(b0+b1*ADB[0])*RF[0]-test_y[0])**2)

bi = [1,1]
sol= minimize(objective, bi, method='SLSQP')
sol

提前致谢。

标签： iterationscipy-optimize

解决方案

我的完整示例：

from sklearn.datasets import make_regression
import pandas as pd
import numpy as np
from scipy.optimize import minimize
from sklearn.ensemble import RandomForestRegressor
from sklearn.ensemble import AdaBoostRegressor
from sklearn.metrics import mean_squared_error
from math import sqrt

# generate 2d regression dataset
X, y = make_regression(n_samples=1100,n_features=2, noise=0.7)
# split into train and test
n_train = 100
train_x, test_x = X[:n_train, :], X[n_train:, :]
train_y, test_y = y[:n_train], y[n_train:]
print(train_x.shape, test_x.shape)


### building the regressors
rgs = []
rfr = RandomForestRegressor(n_estimators=10, random_state=4141, n_jobs=-1)
rfr.fit(train_x, train_y)
print('RF RMSE {score}'.format(score= sqrt(mean_squared_error(test_y, 
rfr.predict(test_x)))))
rgs.append(rfr)

RF = rfr.predict(test_x)

### AdaBoost

ADA = AdaBoostRegressor()
ADA.fit(train_x, train_y)
print('ADA RMSE {score}'.format(score= sqrt(mean_squared_error(test_y, ADA.predict(test_x)))))
rgs.append(ADA)

ADB= ADA.predict(test_x)

# generate random floating point values for w[i]
from numpy.random import seed
from numpy.random import rand
# seed random number generator
seed(511)
# generate random numbers between 0-1
w1 = rand(1000)

仅针对一个数据进行优化：

def objective(b):
    b0 = b[0]
    b1 = b[1]
    return sqrt((w1[0]**(b0+b1*ADB[0])*ADB[0]+(1-w1[0])**(b0+b1*ADB[0])*RF[0]-test_y[0])**2)

bi = [1,0]
sol= minimize(objective, bi, method='SLSQP')
sol

我试图制作循环，但它给出了ValueError: Objective function must return a scalar

loss_es = list()
def objective(b):
    b0 = b[0]
    b1 = b[1]
    for i in range(1000):
        loss = sqrt(((w1[i]**(b0+b1*ADB[i]))*ADB[i]+((1-w1[i])**(b0+b1*ADB[i]))*RF[i]-test_y[i])**2)
        loss_es.append(loss)
    return loss_es

bi = [1,0]

for loss in range(1000):
    sol = minimize(objective, bi, method='SLSQP' )

推荐阅读