**

Tensorflow 机器学习项目实战 记录

---

**

基本操作

---

简单矩阵运算

import tensorflow as tfsess = tf.InteractiveSession()

x = tf.constant([[2, 5, 3, -5],            [0, 3, -2, 5],            [4, 3, 5, 3],            [6, 1, 4, 0]])y = tf.constant([[4, -7, 4, -3, 4],            [6, 4, -7, 4, 7],            [2, 3, 2, 1, 4],            [1, 5, 5, 5, 2]])floatx = tf.constant([[2., 5., 3., -5.],                [0., 3., -2., 5.],                [4., 3., 5., 3.],                [6., 1., 4., 0.]])

  # 矩阵转置tf.transpose(x).eval()  array([[ 2,  0,  4,  6],       [ 5,  3,  3,  1],       [ 3, -2,  5,  4],       [-5,  5,  3,  0]])

#矩阵相乘  Matrix Multiplicationtf.matmul(x, y).eval()array([[ 2,  0,  4,  6],       [ 5,  3,  3,  1],       [ 3, -2,  5,  4],       [-5,  5,  3,  0]])

#矩阵行列式（determinant）tf.matrix_determinant(floatx).eval()817.99969

#逆矩阵(inverse matrix)tf.matrix_inverse(floatx).eval()array([[-0.00855745,  0.10513447, -0.18948655,  0.29584354],       [ 0.12958434,  0.12224938,  0.01222495, -0.05134475],       [-0.01955992, -0.18826404,  0.28117359, -0.18092909],       [-0.08557458,  0.05134474,  0.10513448, -0.0415648 ]], dtype=float32)

#行列式求解tf.matrix_solve(floatx, [[1],[1],[1],[1]]).eval()array([[ 0.202934  ],       [ 0.21271393],       [-0.10757945],       [ 0.02933985]], dtype=float32)

**

约简(reduction)

**

x = tf.constant([[1, 2, 3],                [3, 2, 1],                [-1, -2, -3]])boolean_tensor = tf.constant([[True, False, True],                             [False, False, True],                             [True, False, False]])

# 乘积方式降维，reduction_indices=1在行内计算，reduction_indices=0在列内计算tf.reduce_prod(x, reduction_indices= 1).eval()array([ 6,  6, -6])

tf.reduce_prod(x, reduction_indices= 0).eval(0)Out[15]:array([-3, -8, -9])

# 最小值降维tf.reduce_min(x, reduction_indices=1).eval()array([ 1,  1, -3])

#最大值降维tf.reduce_max(x, reduction_indices=1).eval()array([ 3,  3, -1])

#平均值tf.reduce_mean(x, reduction_indices=1).eval()array([ 2,  2, -2])

# 全部为真，则为真tf.reduce_all(boolean_tensor,reduction_indices=1).eval()array([False, False, False], dtype=bool)

# 存在真，则为真tf.reduce_any(boolean_tensor, reduction_indices=1).eval()array([ True,  True,  True], dtype=bool)

张量分解

seg_ids = tf.constant([0, 1, 1, 2, 2]) # 里面的数字代表分解后所在的index位置。分组的索引： 0 | 1，2 | 3， 4tens1 = tf.constant([[2, 5, 3, -5],                   [0, 3, -2, 5],                   [4, 3, 5, 3],                   [6, 1, 4, 0],                   [6, 1, 4, 0]])

#0组求和作为新的0组  1，2组求和作为新的第1组  3，4组求和作为新的第2组tf.segment_sum(tens1,segment_ids=seg_ids).eval() # sum segmentationarray([[ 2,  5,  3, -5],       [ 4,  6,  3,  8],       [12,  2,  8,  0]])

tf.segment_prod(tens1, segment_ids=seg_ids).eval()array([[  2,   5,   3,  -5],       [  0,   9, -10,  15],       [ 36,   1,  16,   0]])

还有tf.segment_max，tf.segment_min,tf.segment_mean等

序列

x = tf.constant([[2, 5, 3, -5],                [0, 3, -2, 5],                [4, 3, 5, 3],                [6, 1, 4, 0]])listx = tf.constant([1,2,3,4,5,6,7,8])listy = tf.constant([4,5,8,9])boolx = tf.constant([[True, False],                    [False, True]]) # 返回x中最大元素的坐标tf.argmax(x,axis=1).eval()array([1, 3, 2, 0], dtype=int64)

#同理tf.argmin(x,axis=1).eval()array([3, 2, 1, 3], dtype=int64)#tf.listdiff  被移除#返回真值坐标tf.where(boolx).eval()array([[0, 0],       [1, 1]], dtype=int64)

#返回listx中不重复包含的数，并且返回这些数的下标indexitem, index = tf.unique(listx)print(item.eval())print(index.eval())[1 2 3 4 5 6 7 8][0 1 2 3 4 5 6 7]

张量形状变换

x = tf.constant([[2, 5, 3, -5],                [0, 3, -2, 5],                [4, 3, 5, 3],                [6, 1, 4, 0]])tf.shape(x).eval() array([4, 4])

# tensor大小tf.size(x).eval()16#tensor的rank(秩) 等于维度tf.rank(x).eval()2# -1代表视其他维度而定tf.reshape(x , [-1, 2]).eval()array([[ 2,  5],       [ 3, -5],       [ 0,  3],       [-2,  5],       [ 4,  3],       [ 5,  3],       [ 6,  1],       [ 4,  0]])

y = tf.reshape(x, [-1, 1])print(tf.shape(y).eval())y.eval()[16  1]Out[66]:array([[ 2],       [ 5],       [ 3],       [-5],       [ 0],       [ 3],       [-2],       [ 5],       [ 4],       [ 3],       [ 5],       [ 3],       [ 6],       [ 1],       [ 4],       [ 0]]) """Given a tensor `input`, this operation returns a tensor of the same type with    all dimensions of size 1 removed. If you don't want to remove all size 1    dimensions, you can remove specific size 1 dimensions by specifying"""#  维度大小为1的都被移除z = tf.squeeze(y)z.shape(16,)zarray([ 2,  5,  3, -5,  0,  3, -2,  5,  4,  3,  5,  3,  6,  1,  4,  0])

  ```python    # 't' is a tensor of shape [2]    tf.shape(tf.expand_dims(t, 0))  # [1, 2]    tf.shape(tf.expand_dims(t, 1))  # [2, 1]    tf.shape(tf.expand_dims(t, -1))  # [2, 1]    # 't2' is a tensor of shape [2, 3, 5]    tf.shape(tf.expand_dims(t2, 0))  # [1, 2, 3, 5]    tf.shape(tf.expand_dims(t2, 2))  # [2, 3, 1, 5]    tf.shape(tf.expand_dims(t2, 3))  # [2, 3, 5, 1]    ```# 在指定的axis上扩展维度tf.expand_dims(x, 1).eval()array([[[ 2,  5,  3, -5]],       [[ 0,  3, -2,  5]],       [[ 4,  3,  5,  3]],       [[ 6,  1,  4,  0]]])