numpy.core._exceptions.UFuncTypeError: ufunc 'subtract' did not contain a loop with signature matching types (dtype('<U1'), dtype('float64')) -> None

在机器学习实战的Logistic回归梯度上升优化算法中遇到了这个问题

numpy.core._exceptions.UFuncTypeError: ufunc 'subtract' did not contain a loop with signature matching types (dtype('<U1'), dtype('float64')) -> None

代码如下

import math
import numpy as np

def loadDataSet():
    dataSet = []
    labelSet = []
    with open('testSet.txt') as fbj:
        for line in fbj.readlines():
            lineArr = line.strip().split()
            # print(lineArr)
            dataSet.append([1.0, float(lineArr[0]), float(lineArr[1])])
            labelSet.append(lineArr[2])
    return dataSet, labelSet

def sigmoid(inX):
    result = 1/(1+np.exp(-inX))
    return result

def gradAscent(dataSet, labelSet):
    dataMtrix = np.mat(dataSet)
    labelMat = np.mat(labelSet).transpose()
    m, n = np.shape(dataMtrix)
    alpha = 0.001
    maxCycles = 500
    weights = np.ones((n, 1))
    for _ in range(maxCycles):
        h = sigmoid(dataMtrix * weights)
        error = labelMat - h
        weights = weights + alpha * dataMtrix.transpose() * error
    return weights

dataSet, labelSet = loadDataSet()
# print(dataSet)
# print(labelSet)
print(gradAscent(dataSet, labelSet))

这里报错说的是数据类型不符不能相减

那么分别查看一下(在jupyter调试)

labelMat.dtype

dtype('<U1')

h.dtype

dtype('float64')

那么解决办法就是将<U1类型换成float64

但是使用如下方法还是报错

labelMat.dtype = 'float64'

ValueError: When changing to a larger dtype, its size must be a divisor of the total size in bytes of the last axis of the array.

那么只好乖乖使用astype方法

labelMat = labelMat.astype(np.float64)

修改后的代码及结果如下

import math
import numpy as np

def loadDataSet():
    dataSet = []
    labelSet = []
    with open('testSet.txt') as fbj:
        for line in fbj.readlines():
            lineArr = line.strip().split()
            # print(lineArr)
            dataSet.append([1.0, float(lineArr[0]), float(lineArr[1])])
            labelSet.append(lineArr[2])
    return dataSet, labelSet

def sigmoid(inX):
    result = 1/(1+np.exp(-inX))
    return result

def gradAscent(dataSet, labelSet):
    dataMtrix = np.mat(dataSet)
    labelMat = np.mat(labelSet).transpose()
    labelMat = labelMat.astype(np.float64)
    m, n = np.shape(dataMtrix)
    alpha = 0.001
    maxCycles = 500
    weights = np.ones((n, 1))
    for _ in range(maxCycles):
        h = sigmoid(dataMtrix * weights)
        error = labelMat - h
        weights = weights + alpha * dataMtrix.transpose() * error
    return weights

dataSet, labelSet = loadDataSet()
# print(dataSet)
# print(labelSet)
print(gradAscent(dataSet, labelSet))

[[ 4.12414349]
 [ 0.48007329]
 [-0.6168482 ]]