广场遇到溢出

Question

我尝试搜索我遇到的溢出错误，但没有成功。

当我运行这个程序时，我得到运行对我来说毫无意义的时间错误。这是我使用的数据：https://pastebin.com/MLWvUarm

import numpy as np

def loadData():
    data = np.loadtxt('data.txt', delimiter=',')
    x = np.c_[data[:,0:2]]
    y = np.c_[data[:,-1]]
    return x, y

def hypothesis(x, theta):
    h = x.dot(theta)
    return h

def computeCost(x, y, theta):
    m = np.size(y, 0)
    h = hypothesis(x, theta)
    J = (1/(2*m)) * np.sum(np.square(h-y))
    return J

def gradient_descent(x, y, theta, alpha, mxIT):
    m = np.size(y, 0)
    J_history = np.zeros((mxIT, 1))

    for it in range(mxIT):
        hyp = hypothesis(x, theta)
        err = hyp - y
        theta = theta - (alpha/m) * (x.T.dot(err))
        J_history[it] = computeCost(x, y, theta)

    return theta, J_history

def main():
    x, y = loadData()
    x = np.c_[np.ones(x.shape[0]), x]
    theta = np.zeros((np.size(x, 1), 1))
    alpha = 0.01
    mxIT = 400

    theta, j_his = gradient_descent(x, y, theta, alpha, mxIT)
    print(theta)

if __name__ == "__main__":
    main()

如何解决这个问题？

Answer 1

加载x后，尝试除以均值，看是否收敛。 Link 均值文档：numpy.mean

...
x, y = loadData()
x = x / x.mean(axis=0, keepdims=True)
x = np.c_[np.ones(x.shape[0]), x]
...

目前它似乎有分歧，这会产生非常高的错误，numpy 会抱怨。您可以从您在 J_history.

中维护的成本历史记录中看到这一点

广场遇到溢出

Overflow encountered in square

python

numpy

overflow