numpy创建单位矩阵和对角矩阵的实例

在学习linear regression时经常处理的数据一般多是矩阵或者n维向量的数据形式，所以必须对矩阵有一定的认识基础。

numpy中创建单位矩阵借助identity（）函数。更为准确的说，此函数创建的是一个n*n的单位数组，返回值的dtype=array数据形式。其中接受的参数有两个，第一个是n值大小，第二个为数据类型，一般为浮点型。单位数组的概念与单位矩阵相同，主对角线元素为1，其他元素均为零，等同于单位1。而要想得到单位矩阵，只要用mat（）函数将数组转换为矩阵即可。

>>> import numpy as np
>>> help(np.identity)

Help on function identity in module numpy:

identity(n, dtype=None)
 Return the identity array.

The identity array is a square array with ones on
 the main diagonal.

Parameters
 ----------
 n : int
   Number of rows (and columns) in `n` x `n` output.
 dtype : data-type, optional
   Data-type of the output. Defaults to ``float``.

Returns
 -------
 out : ndarray
   `n` x `n` array with its main diagonal set to one,
   and all other elements 0.

Examples
 --------
 >>> np.identity(3)
 array([[ 1., 0., 0.],
     [ 0., 1., 0.],
     [ 0., 0., 1.]])
>>> np.identity(5)

array([[1., 0., 0., 0., 0.],
   [0., 1., 0., 0., 0.],
   [0., 0., 1., 0., 0.],
   [0., 0., 0., 1., 0.],
   [0., 0., 0., 0., 1.]])
>>> A = np.mat(np.identity(5))

>>> A

matrix([[1., 0., 0., 0., 0.],
   [0., 1., 0., 0., 0.],
   [0., 0., 1., 0., 0.],
   [0., 0., 0., 1., 0.],
   [0., 0., 0., 0., 1.]])

矩阵的运算中还经常使用对角阵，numpy中的对角阵用eye（）函数来创建。eye（）函数接受五个参数，返回一个单位数组。第一个和第二个参数N，M分别对应表示创建数组的行数和列数，当然当你只设定一个值时，就默认了N=M。第三个参数k是对角线指数，跟diagonal中的offset参数是一样的，默认值为0，就是主对角线的方向，上三角方向为正，下三角方向为负，可以取-n到+m的范围。第四个参数是dtype，用于指定元素的数据类型，第五个参数是order，用于排序，有‘C'和‘F'两个参数，默认值为‘C'，为行排序，‘F'为列排序。返回值为一个单位数组。

>>> help(np.eye)

Help on function eye in module numpy:

eye(N, M=None, k=0, dtype=<class 'float'>, order='C')
 Return a 2-D array with ones on the diagonal and zeros elsewhere.

Parameters
 ----------
 N : int
  Number of rows in the output.
 M : int, optional
  Number of columns in the output. If None, defaults to `N`.
 k : int, optional
  Index of the diagonal: 0 (the default) refers to the main diagonal,
  a positive value refers to an upper diagonal, and a negative value
  to a lower diagonal.
 dtype : data-type, optional
  Data-type of the returned array.
 order : {'C', 'F'}, optional
   Whether the output should be stored in row-major (C-style) or
   column-major (Fortran-style) order in memory.

.. versionadded:: 1.14.0

Returns
 -------
 I : ndarray of shape (N,M)
  An array where all elements are equal to zero, except for the `k`-th
  diagonal, whose values are equal to one.

See Also
 --------
 identity : (almost) equivalent function
 diag : diagonal 2-D array from a 1-D array specified by the user.

Examples
 --------
 >>> np.eye(2, dtype=int)
 array([[1, 0],
     [0, 1]])
 >>> np.eye(3, k=1)
 array([[ 0., 1., 0.],
     [ 0., 0., 1.],
     [ 0., 0., 0.]])

numpy中的diagonal（）方法可以对n*n的数组和方阵取对角线上的元素，diagonal（）接受三个参数。第一个offset参数是主对角线的方向，默认值为0是主对角线，上三角方向为正，下三角方向为负，可以取-n到+n的范围。第二个参数和第三个参数是在数组大于2维时指定一个2维数组时使用，默认值axis1=0，axis2=1。

>>> help(A.diagonal)

Help on built-in function diagonal:

diagonal(...) method of numpy.matrix instance
 a.diagonal(offset=0, axis1=0, axis2=1)

Return specified diagonals. In NumPy 1.9 the returned array is a
 read-only view instead of a copy as in previous NumPy versions. In
 a future version the read-only restriction will be removed.

Refer to :func:`numpy.diagonal` for full documentation.

See Also
 --------
 numpy.diagonal : equivalent function
>>> help(np.diagonal)

Help on function diagonal in module numpy:

diagonal(a, offset=0, axis1=0, axis2=1)
 Return specified diagonals.

If `a` is 2-D, returns the diagonal of `a` with the given offset,
 i.e., the collection of elements of the form ``a[i, i+offset]``. If
 `a` has more than two dimensions, then the axes specified by `axis1`
 and `axis2` are used to determine the 2-D sub-array whose diagonal is
 returned. The shape of the resulting array can be determined by
 removing `axis1` and `axis2` and appending an index to the right equal
 to the size of the resulting diagonals.

In versions of NumPy prior to 1.7, this function always returned a new,
 independent array containing a copy of the values in the diagonal.

In NumPy 1.7 and 1.8, it continues to return a copy of the diagonal,
 but depending on this fact is deprecated. Writing to the resulting
 array continues to work as it used to, but a FutureWarning is issued.

Starting in NumPy 1.9 it returns a read-only view on the original array.
 Attempting to write to the resulting array will produce an error.

In some future release, it will return a read/write view and writing to
 the returned array will alter your original array. The returned array
 will have the same type as the input array.

If you don't write to the array returned by this function, then you can
 just ignore all of the above.

If you depend on the current behavior, then we suggest copying the
 returned array explicitly, i.e., use ``np.diagonal(a).copy()`` instead
 of just ``np.diagonal(a)``. This will work with both past and future
 versions of NumPy.

Parameters
 ----------
 a : array_like
   Array from which the diagonals are taken.
 offset : int, optional
   Offset of the diagonal from the main diagonal. Can be positive or
   negative. Defaults to main diagonal (0).
 axis1 : int, optional
   Axis to be used as the first axis of the 2-D sub-arrays from which
   the diagonals should be taken. Defaults to first axis (0).
 axis2 : int, optional
   Axis to be used as the second axis of the 2-D sub-arrays from
   which the diagonals should be taken. Defaults to second axis (1).

Returns
 -------
 array_of_diagonals : ndarray
   If `a` is 2-D, then a 1-D array containing the diagonal and of the
   same type as `a` is returned unless `a` is a `matrix`, in which case
   a 1-D array rather than a (2-D) `matrix` is returned in order to
   maintain backward compatibility.

If ``a.ndim > 2``, then the dimensions specified by `axis1` and `axis2`
   are removed, and a new axis inserted at the end corresponding to the
   diagonal.

Raises
 ------
 ValueError
   If the dimension of `a` is less than 2.

See Also
 --------
 diag : MATLAB work-a-like for 1-D and 2-D arrays.
 diagflat : Create diagonal arrays.
 trace : Sum along diagonals.

Examples
 --------
 >>> a = np.arange(4).reshape(2,2)
 >>> a
 array([[0, 1],
     [2, 3]])
 >>> a.diagonal()
 array([0, 3])
 >>> a.diagonal(1)
 array([1])

A 3-D example:

>>> a = np.arange(8).reshape(2,2,2); a
 array([[[0, 1],
     [2, 3]],
     [[4, 5],
     [6, 7]]])
 >>> a.diagonal(0, # Main diagonals of two arrays created by skipping
 ...      0, # across the outer(left)-most axis last and
 ...      1) # the "middle" (row) axis first.
 array([[0, 6],
     [1, 7]])

The sub-arrays whose main diagonals we just obtained; note that each
 corresponds to fixing the right-most (column) axis, and that the
 diagonals are "packed" in rows.

>>> a[:,:,0] # main diagonal is [0 6]
 array([[0, 2],
     [4, 6]])
 >>> a[:,:,1] # main diagonal is [1 7]
 array([[1, 3],
     [5, 7]])
>>> A = np.random.randint(low=5, high=30, size=(5, 5))

>>> A

array([[25, 15, 26, 6, 22],
   [27, 14, 22, 16, 21],
   [22, 17, 10, 14, 25],
   [11, 9, 27, 20, 6],
   [24, 19, 19, 26, 14]])
>>> A.diagonal()

array([25, 14, 10, 20, 14])
>>> A.diagonal(offset=1)

array([15, 22, 14, 6])
>>> A.diagonal(offset=-2)

array([22, 9, 19])