numpy实用技巧（一） python

import numpy as np

功能	函数
设置打印输出的精度	np.set_printoptions(precision=4)

一维数组构成的list，再进行像数组的转换是以行的形式（而非列的形式）在拼接；

>>> X = np.random.randn(5, 3) >>> ops = [np.argmax, np.argmin] >>> np.asarray([op(X, 1) for op in ops]) [[1 2 2 2 2] [2 0 1 1 1]]

两个向量的拼接例如两个长度为n的一维向量，拼接为2×n的矩阵，可以使用np.vstack()，

>>> x, y = np.ones(3), np.zeros(3) >>> np.vstack((x, y)) array([[ 1.,1.,1.], [ 0.,0.,0.]])

【numpy实用技巧（一）】也可使用作为np.array()构造函数的参数，进行创建：

>>> np.array([x, y]) array([[ 1.,1.,1.], [ 0.,0.,0.]])

xx, yy = np.meshgrid(np.arange(, , ), np.arange(, , )) z = clf.predict(np.array([xx.ravel(), yy.ravel()]).T) z = z.reshape(xx.shape) plt.contoutf(xx, yy, z, alpha=.4, cmap=cmap)

numpy.ndarray的遍历

>>> X = np.random.randn(3, 3) array([[-0.24882132, -0.32389773, -0.96069467], [ 1.26331248,1.59089579, -0.97145676], [-0.03989954,0.28587614,0.04657364]])>>> for i in X: print(i) [-0.24882132 -0.32389773 -0.96069467] [ 1.263312481.59089579 -0.97145676] [-0.039899540.285876140.04657364]>>> for i in X: for j in i: print j-0.248821323982 -0.32389773407 -0.960694672326 1.26331248229 1.59089578902 -0.971456755866 -0.0398995441063 0.285876139182 0.0465736443469

numpy.ndarray()类型转换有时编译器会报如下的错误：

TypeError: Cannot cast array data from dtype(‘float64’) to
dtype(‘int32’) according to the rule ‘safe’

>>> print(np.bincount(y_train))TypeError: Cannot cast array data from dtype('float64') to dtype('int32') according to the rule 'safe'

如果我们使用python基本模块下的强制类型转换，又会提示如下的错误：

>>> print(np.bincount(int(y_train)))TypeError: only length-1 arrays can be converted to Python scalars

此时可以使用x.astype(type)成员

>>> print(np.bincount(y_train.astype(np.int32))

二维数组的逆序列向的逆序 A[:, -1::-1]：

>>> np.set_printoptions(precision=4) >>> A = np.random.randn(3, 3) >>> A array([[ 1.2381, -0.2428, -0.4687], [-1.0588,0.0432,0.9937], [ 0.2708,1.4833,0.2697]])>>> A[:, -1::-1] array([[-0.4687, -0.2428,1.2381], [ 0.9937,0.0432, -1.0588], [ 0.2697,1.4833,0.2708]])

行向的逆序：A[-1::-1, :]