TensorFlow2.1 Basic concepts and common functions

One 、 Basic concepts

1.1 Tensor The concept of

TensorFlow Medium Tensor It's a tensor , It's a multidimensional array （ list ）, Dimension of tensor expressed by order .

dimension	rank	name	give an example
0-D	0	Scalar scalar	$s=1,2,3$
1-D	1	vector vector	$v=[1,2,3]$
2-D	2	matrix matrix	$m=[[1,2,3]],[4,5,6]],[7,8,9]]$
n-D	3	tensor tensor	$t=[[[...]]]nindividual$

Tensors can represent 0 Step to n Order array （ list ）

1.2 Tensor data type

• tf.int,tf.float integer 、 floating-point
  tf.int 32,tf.float 32,tf.float 64
• tf.bool Boolean type
  tf.constant([True,False])
• tf.string String type
  tf.constant(“Hello,world!”)

1.3 establish Tensor

1. Create a tensor directly

tf.constant( Tensor content ,dtype= data type ( Optional ))

import tensorflow as tf
a=tf.constant([1,5],dtype=tf.int64)
print(a)
print(a.dtype)
print(a.shape)

The operation results are as follows ：

<tf.Tensor([1,5], shape=(2 , ) , dtype=int64)
<dtype: 'int64'>
(2,)

2. take numpy The data type of is converted to Tensor data type

tf.convert_to_tensor( Data name ,dtype= data type ( Optional ))

import tensorflow as tf
import numpy as np
a = np.arange(0, 5)
b = tf.convert_to_tensor( a, dtype=tf.int64 )
print(a)
print(b)

The operation results are as follows ：

[0 1 2 3 4]
tf.Tensor([0 1 2 3 4], shape=( 5 , ), dtype=int64)

3. Use the function to initialize

• Create all for 0 Tensor tf.zeros( dimension )
• Create all for 1 Tensor tf.ones( dimension )
• Create tensors with all specified values tf.fill( dimension , Specify the value )

dimension ：

• A one-dimensional Write the number directly
• A two-dimensional use $[That\text{'}s\; ok\; ,\; Column]$
• Multidimensional use $[n,m,j,k\dots \dots ]$

a = tf.zeros([2, 3])
b = tf.ones(4)
c = tf.fill([2, 2], 9)
print(a)
print(b)
print(c)

The operation results are as follows ：

tf.Tensor([[0. 0. 0.] [0. 0. 0.]], shape=(2, 3), dtype=float32)
tf.Tensor([1. 1. 1. 1.], shape=(4, ), dtype=float32)
tf.Tensor([[9 9] [9 9]], shape=(2, 2), dtype=int32)

• Generating random number of normal distribution , The default mean is 0, The standard deviation is 1
  tf. random.normal ( dimension ,mean= mean value ,stddev= Standard deviation )
• Generate the random number of truncated normal distribution
  tf. random.truncated_normal ( dimension ,mean= mean value ,stddev= Standard deviation )
  stay tf.truncated_normal If the value of randomly generated data in （μ-2σ,μ+2σ） outside , Then regenerate , Ensure that the generated value is near the mean value .

d = tf.random.normal ([2, 2], mean=0.5, stddev=1)
print(d)
e = tf.random.truncated_normal ([2, 2], mean=0.5, stddev=1)
print(e)

The operation results are as follows ：

tf.Tensor(
[[0.7925745 0.643315 ]
[1.4752257 0.2533372]], shape=(2, 2), dtype=float32)
tf.Tensor(
[[ 1.3688478 1.0125661 ]
[ 0.17475659 -0.02224463]], shape=(2, 2), dtype=float32)

• Generate evenly distributed random numbers ( minval, maxval )
  tf.random.uniform( dimension ,minval= minimum value ,maxval= Maximum )

f = tf.random.uniform([2, 2], minval=0, maxval=1)
print(f)

tf.Tensor(
[[0.28219545 0.15581512]
[0.77972126 0.47817433]], shape=(2, 2), dtype=float32)

Two 、 Common functions

2.1 Statistical class functions

1. mandatory tensor Convert to the data type
   tf.cast ( Zhang Liangming ,dtype= data type )
2. Calculate the minimum value of the element in the tensor dimension
   tf.reduce_min ( Zhang Liangming )
3. Calculate the maximum value of the element in the tensor dimension
   tf.reduce_max ( Zhang Liangming )

x1 = tf.constant ([1., 2., 3.],
dtype=tf.float64)
print(x1)
x2 = tf.cast (x1, tf.int32)
print(x2)
print (tf.reduce_min(x2), 
tf.reduce_max(x2))

The operation results are as follows ：

tf.Tensor([1. 2. 3.], shape=(3,), dtype=float64)
tf.Tensor([1 2 3], shape=(3,), dtype=int32)
tf.Tensor(1, shape=(), dtype=int32)
tf.Tensor(3, shape=(), dtype=intt32)

In a two-dimensional tensor or array , Can be adjusted by axis be equal to 0 or 1 Control execution dimension .

• axis=0 On behalf of inter-bank （ longitude ,down),
• axis=1 For cross column （ latitude ,across)
• If you don't specify axis, Then all elements participate in the calculation

4. Calculates the mean value of the tensor along the specified dimension
tf.reduce_mean ( Zhang Liangming ,axis= Operating shaft )
5. Calculate the sum of the tensor along the specified dimension
tf.reduce_sum ( Zhang Liangming ,axis= Operating shaft )

x=tf.constant( [ [ 1, 2, 3],[ 2, 2, 3] ] )
print(x)
print(tf.reduce_mean( x ))
print(tf.reduce_sum( x, axis=1 ))

The operation results are as follows ：

tf.Tensor([[1 2 3][2 2 3]], shape=(2, 3), dtype=int32)
tf.Tensor(2, shape=(), dtype=int32)
tf.Tensor([6 7], shape=(2,), dtype=int32)

2.2 Mathematical operation function

2.2.1 arithmetic

1. Add the corresponding elements of two tensors
   tf.add ( tensor 1, tensor 2)
2. Subtracting the corresponding elements of two tensors
   tf.subtract ( tensor 1, tensor 2)
3. Multiplication of the corresponding elements of two tensors
   tf.multiply ( tensor 1, tensor 2)
4. To divide the corresponding elements of two tensors
   tf.divide ( tensor 1, tensor 2)

Only Same dimension The tensor of can do four operations

a = tf.ones([1, 3])
b = tf.fill([1, 3], 3.)
print(a)
print(b)
print(tf.add(a,b))
print(tf.subtract(a,b))
print(tf.multiply(a,b))
print(tf.divide(b,a))

The operation results are as follows ：

tf.Tensor([[1. 1. 1.]], shape=(1, 3), dtype=float32)
tf.Tensor([[3. 3. 3.]], shape=(1, 3), dtype=float32
tf.Tensor([[4. 4. 4.]], shape=(1, 3), dtype=float32)
tf.Tensor([[-2. -2. -2.]], shape=(1, 3), dtype=float32)
tf.Tensor([[3. 3. 3.]], shape=(1, 3), dtype=float32)
tf.Tensor([[3. 3. 3.]], shape=(1, 3), dtype=float32)

2.2.2 square 、 Power and square

1. Calculate the square of a tensor ：tf.square ( Zhang Liangming )
2. To compute a tensor n Power ：tf.pow ( Zhang Liangming ,n Power number )
3. Calculate the square root of a tensor ：tf.sqrt ( Zhang Liangming ）

a = tf.fill([1, 2], 3.)
print(a)
print(tf.pow(a, 3))
print(tf.square(a))
print(tf.sqrt(a))

The operation results are as follows ：

tf.Tensor([[3. 3.]], shape=(1, 2), dtype=float32)
tf.Tensor([[27. 27.]], shape=(1, 2), dtype=float32)
tf.Tensor([[9. 9.]], shape=(1, 2), dtype=float32)
tf.Tensor([[1.7320508 1.7320508]], shape=(1, 2), dtype=float32)

2.2.3 Matrix multiplication

Realize the multiplication of two matrices ：tf.matmul( matrix 1, matrix 2)

a = tf.ones([3, 2])
b = tf.fill([2, 3], 3.)
print(tf.matmul(a, b))

The operation results are as follows ：

tf.Tensor(
[[6. 6. 6.]
[6. 6. 6.]
[6. 6. 6.]], shape=(3, 3), dtype=float32)

2.3 Other common functions

1. tf.Variable()

tf.Variable () Mark the variable as “ Can be trained ”, The marked variable propagates in the back
Record gradient information in . In neural network training , This function is often used to mark the parameters to be trained

w = tf.Variable(tf.random.normal([2, 2], mean=0, stddev=1))

2.tf.data.Dataset.from_tensor_slices()

Segment the first dimension of the incoming tensor , Generate input features / Label pair , Build data set
data = tf.data.Dataset.from_tensor_slices(( Input characteristics , label ))

features = tf.constant([12,23,10,17])
labels = tf.constant([0, 1, 1, 0])
dataset = tf.data.Dataset.from_tensor_slices((features, labels))
print(dataset)
for element in dataset:
   print(element)

The operation results are as follows ：

<TensorSliceDataset shapes: ((),()), types: (tf.int32, tf.int32))> 
(<tf.Tensor: id=9, shape=(), dtype=int32, numpy=12>, <tf.Tensor: id=10, shape=(), dtype=int32, numpy=0>) 
(<tf.Tensor: id=11, shape=(), dtype=int32, numpy=23>, <tf.Tensor: id=12, shape=(), dtype=int32, numpy=1>) 
(<tf.Tensor: id=13, shape=(), dtype=int32, numpy=10>, <tf.Tensor: id=14, shape=(), dtype=int32, numpy=1>) 
(<tf.Tensor: id=15, shape=(), dtype=int32, numpy=17>, <tf.Tensor: id=16, shape=(), dtype=int32, numpy=0>)

3.tf.GradientTape()

with The structure records the calculation process ,gradient Find the gradient of the tensor

with tf.GradientTape( ) as tape:
   w = tf.Variable(tf.constant(3.0))
   loss = tf.pow(w,2) #loss=w^2 loss’=2w
grad = tape.gradient(loss,w) 
print(grad)

The operation results are as follows ：

tf.Tensor(6.0, shape=(), dtype=float32)

4.enumerate()

enumerate yes python Is a built-in function of , It can traverse every element ( As listing 、 Tuples
Or a string ), Combination for ： Indexes Elements , Often in for Use in loop .

seq = ['one', 'two', 'three']
for i, element in enumerate(seq):
   print(i, element)

The operation results are as follows ：

0 one
1 two
2 three

5.tf.one_hot()

Hot coding alone （one-hot encoding）： In the classification problem , It is often labeled with a unique heat code ,
Tag category ：1 Said is ,0 Express non
tf.one_hot() Function to convert data , Convert to one-hot Form of data output

tf.one_hot ( Data to be converted , depth= Several categories )

classes = 3
labels = tf.constant([1,0,2]) #  The minimum value of the input element is 0, The maximum is 2
output = tf.one_hot( labels, depth=classes )
print(output)

The operation results are as follows ：

([[0. 1. 0.]
[1. 0. 0.]
[0. 0. 1.]], shape=(3, 3), dtype=float32)

6.tf.nn.softmax()

$$Softmax(y_i)=\frac{e^{y_i}}{{\sum }_{j=0}^n{e}^{y_j}}$$
When n Classified n Outputs $（y_0,y_1,\dots \dots y_{n-1}）$ adopt softmax( ) function ,
It conforms to the probability distribution
$$\forall P(X=x)\in [0,1]And\sum _{x}P(X=x)=1$$

y = tf.constant ( [1.01, 2.01, -0.66] )
y_pro = tf.nn.softmax(y)
print("After softmax, y_pro is:", y_pro)

The operation results are as follows ：

After softmax, y_pro is: tf.Tensor([0.25598174 0.69583046 0.0481878], shape=(3,), dtype=float32)

7.assign_sub()

Assignment operation , Update the value of the parameter and return .
call assign_sub front , First use tf.Variable Defining variables w To be able to train （ Self updating ）.

w.assign_sub (w Content to be subtracted )

w = tf.Variable(4)
w.assign_sub(1)
print(w)

<tf.Variable 'Variable:0' shape=() dtype=int32, numpy=3>

8.tf.argmax()

Returns the index of the tensor along the maximum value of the specified dimension

tf.argmax ( Zhang Liangming ,axis= Operating shaft )

import numpy as np
test = np.array([[1, 2, 3], [2, 3, 4], [5, 4, 3], [8, 7, 2]])
print(test)
print( tf.argmax (test, axis=0)) #  Return each column （ longitude ） Index of maximum value 
print( tf.argmax (test, axis=1)) #  Return each line （ latitude ） Index of maximum value 

The operation results are as follows ：

[[1 2 3]
[2 3 4]
[5 4 3]
[8 7 2]]
tf.Tensor([3 3 1], shape=(3,), dtype=int64)
tf.Tensor([2 2 0 0], shape=(4,), dtype=int64)

9.tf.where()

tf.where( Conditional statements , Really back to A, False return B)

a=tf.constant([1,2,3,1,1])
b=tf.constant([0,1,3,4,5])
c=tf.where(tf.greater(a,b), a, b) #  if a>b, return a Elements corresponding to the position , otherwise 
 return b Elements corresponding to the position 
print("c:",c)

The operation results are as follows ：

c： tf.Tensor([1 2 3 4 5], shape=(5,), dtype=int32)

10.np.random.RandomState.rand()

Return to one [0,1) Random number between
np.random.RandomState.rand( dimension ) # Dimension is empty , Return scalar

import numpy as np
rdm=np.random.RandomState(seed=1) #seed= The constant generates the same random number every time 
a=rdm.rand() #  Returns a random scalar 
b=rdm.rand(2,3) #  The return dimension is 2 That's ok 3 Column random number matrix 
print("a:",a)
print("b:",b)

The operation results are as follows ：

a: 0.417022004702574
b: [[7.20324493e-01 1.14374817e-04 3.02332573e-01]
[1.46755891e-01 9.23385948e-02 1.86260211e-01]]

11.np.vstack()

Stack the two arrays vertically
np.vstack( Array 1, Array 2)

import numpy as np
a = np.array([1,2,3])
b = np.array([4,5,6])
c = np.vstack((a,b)) 
print("c:\n",c)

The operation results are as follows ：

c:
[[1 2 3]
[4 5 6]]

12.np.mgrid[ ]

np.mgrid[ Starting value : End value : step , Starting value : End value : step , … ]
x.ravel( )： take x To a one-dimensional array
np.c_[ Array 1, Array 2, … ]： Pairing the returned interval values

import numpy as np
x, y = np.mgrid [1:3:1, 2:4:0.5] 
grid = np.c_[x.ravel(), y.ravel()]
print("x:",x)
print("y:",y)
print('grid:\n', grid)

The operation results are as follows ：

x = [[1. 1. 1. 1.]
[2. 2. 2. 2.]]
y = [
[2. 2.5 3. 3.5]
[2. 2.5 3. 3.5]]
grid:
[[1. 2. ]
[1. 2.5]
[1. 3. ]
[1. 3.5]
[2. 2. ]
[2. 2.5]
[2. 3. ]
[2. 3.5]]