scipy优化定义初始猜测的问题

Question

I guess I'm not right at defining x0 (Initial guess)

With minimize from scipy.optimize I looking for min of my fun error and I want 𝑤0 to be in bettween [-100,100] and 𝑤1 in [-5, 5]. Starting point– ( 𝑤0 , 𝑤1 ) = (0, 0).

Function

def error(w0, w1):
    dataset = data
    total_error = 0
    for i in range(1, 10):
        total_error = total_error + (dataset['Height'][i] - (w0 + w1 * dataset['Weight'][i]))**2
    return total_error

Problem

bnds = ((-100, 100), (-5, 5))
x0 = (0), (0)
res = minimize(error, (0), (0), method = "L-BFGS-B", bounds = bnds)

ValueError: length of x0 != length of bounds

WTF

print (len(x0))
print (len(bnds))

2 2

Answer 1

Hey i reproduced your error, i had also some issues here. But i think i found the main problem:

from scipy.optimize import minimize
import numpy as np

def error(w0,w1):
    dataset = np.array([1,2,3,4,5,6,777,7,77,7,7,7])
    total_error = 0
    for i in range(1, 10):
        total_error = total_error + (dataset[i] - (w0+ w1 * dataset[i]))**2
    return total_error


x0 = np.array([[12],[22]])

res =minimize(error, x0[0],x0[1], method = "L-BFGS-B",bounds=[[0,10]])

You just can choose 2 boundarys, because your optimization problem is 2 dimensional. You define your boundaries for your "x-axis" over the input Data points, and you just can maybe wrapp or crop the output to your boundary "bnds". Hope i could help