tensorflow-related APIs

tensorflow-Correlation Apitensorflow Correlation function understanding

Task Time: Unknown time

Tf.truncated_normal

truncated_normal(    shape,    mean=0.0,    stddev=1.0,    dtype=tf.float32,    seed=None,    name=None)

Function Description:

Produces a truncated normal distribution random number, the value range is [mean - 2 * stddev, mean + 2 * stddev] .

Parameter list:

Name of parameter	must-Choose	type	Description
Shape	Is	1-dimensional shaping tensor or array	Dimensions of the output tensor
Mean	Whether	0 dimensional tensor or value	Mean value
StdDev	Whether	0 dimensional tensor or value	Standard deviation
Dtype	Whether	Dtype	Output type
Seed	Whether	Numerical	Random seed, if seed is assigned, produces the same random number each time
Name	Whether	String	Operation name

Example code:

You can now create the source file truncated_normal.pyin the /home/ubuntu directory:

Example code:/home/ubuntu/truncated_normal.py

#!/usr/bin/pythonimport tensorflow as tfinitial = tf.truncated_normal(shape=[3,3], mean=0, stddev=1)print tf.Session().run(initial)

Then execute:

python /home/ubuntu/truncated_normal.py

Execution Result:

Will get a value range [-2, 2] of the 3 * 3 matrix, you can also try to modify the source code to see what changes in the output results?

Tf.constant

constant(    value,    dtype=None,    shape=None,    name=‘Const‘,    verify_shape=False)

Function Description:

Generates a constant tensor of a shape dimension based on value values

Parameter list:

Name of parameter	must-Choose	type	Description
Value	Is	Constant Value or list	The value of the output tensor
Dtype	Whether	Dtype	Output tensor element type
Shape	Whether	1-dimensional shaping tensor or array	Dimensions of the output tensor
Name	Whether	String	Tensor name
Verify_shape	Whether	Boolean	Detects if shape is the same shape as value, and if it is fasle, the last element will be used to complete the shape

Example code:

You can now create the source file constant.pyin the /home/ubuntu directory, which can be referenced in the following:

Example code:/home/ubuntu/constant.py

#!/usr/bin/pythonimport tensorflow as tfimport numpy as npa = tf.constant([1,2,3,4,5,6],shape=[2,3])b = tf.constant(-1,shape=[3,2])c = tf.matmul(a,b)e = tf.constant(np.arange(1,13,dtype=np.int32),shape=[2,2,3])f = tf.constant(np.arange(13,25,dtype=np.int32),shape=[2,3,2])g = tf.matmul(e,f)with tf.Session() as sess:    print sess.run(a)    print ("##################################")    print sess.run(b)    print ("##################################")    print sess.run(c)    print ("##################################")    print sess.run(e)    print ("##################################")    print sess.run(f)    print ("##################################")    print sess.run(g)

Then execute:

python /home/ubuntu/constant.py

Execution Result:

• a:2x3 dimension tensor;
• b:3x2 dimension tensor;
• c:2x2 dimension tensor;
• e:2x2x3 dimension tensor;
• f:2x3x2 dimension tensor;
• g:2x2x2 dimension tensor.

You can also try to modify the source code to see what happens to the output?

Tf.placeholder

placeholder(    dtype,    shape=None,    name=None)

Function Description:

is a placeholder that needs to be supplied with data at execution time

Parameter list:

Name of parameter	must-Choose	type	Description
Dtype	Is	Dtype	Placeholder data type
Shape	Whether	1-dimensional shaping tensor or array	Placeholder Dimension
Name	Whether	String	Placeholder Name

Example code:

You can now create the source file placeholder.pyin the /home/ubuntu directory, which can be referenced in the following:

Example code:/home/ubuntu/placeholder.py

#!/usr/bin/pythonimport tensorflow as tfimport numpy as npx = tf.placeholder(tf.float32,[None,10])y = tf.matmul(x,x)with tf.Session() as sess:    rand_array = np.random.rand(10,10)    print sess.run(y,feed_dict={x:rand_array})

Then execute:

python /home/ubuntu/placeholder.py

Execution Result:

Outputs the tensor of a 10x10 dimension. You can also try to modify the source code to see what happens to the output?

Tf.nn.bias_add

bias_add(    value,    bias,    data_format=None,    name=None)

Function Description:

Adding the deviation item bias to value can be seen as a special case of tf.add where the bias must be one-dimensional and the same as the last dimension of value, and the data type must be the same as value.

Parameter list:

Name of parameter	must-Choose	type	Description
Value	Is	Tensor	Data types are float, double, Int64, Int32, Uint8, Int16, int8, complex64, or complex128
Bias	Is	1-dimensional tensor	Dimensions must be equal to the last dimension of value
Data_format	Whether	String	Data format, supporting ' NHWC ' and ' NCHW '
Name	Whether	String	Operation name

Example code:

You can now create the source file bias_add.pyin the /home/ubuntu directory, which can be referenced in the following:

Example code:/home/ubuntu/bias_add.py

#!/usr/bin/pythonimport tensorflow as tfimport numpy as npa = tf.constant([[1.0, 2.0],[1.0, 2.0],[1.0, 2.0]])b = tf.constant([2.0,1.0])c = tf.constant([1.0])sess = tf.Session()print sess.run(tf.nn.bias_add(a, b)) #print sess.run(tf.nn.bias_add(a,c)) errorprint ("##################################")print sess.run(tf.add(a, b))print ("##################################")print sess.run(tf.add(a, c))

Then execute:

python /home/ubuntu/bias_add.py

Execution Result:

3 x 3x2 dimensions. You can also try to modify the source code to see what happens to the output?

Tf.reduce_mean

reduce_mean(    input_tensor,    axis=None,    keep_dims=False,    name=None,    reduction_indices=None)

Function Description:

Calculate tensor input_tensor Average

Parameter list:

	required	type	description
Input_tensor	Is	Tensor	Enter the tensor of the average to be averaged
Axis	Whether	None, 0, 1	None: Global averaging; 0: averaging of each column; 1: Averaging each row
Keep_dims	Whether	Boolean	Keep the original dimension, down to 1
Name	Whether	String	Operation name
Reduction_indices	Whether	None	is equivalent to axis and is deprecated

Example code:

You can now create the source file reduce_mean.pyin the /home/ubuntu directory, which can be referenced in the following:

Example code:/home/ubuntu/reduce_mean.py

#!/usr/bin/pythonimport tensorflow as tfimport numpy as npinitial = [[1.,1.],[2.,2.]]x = tf.Variable(initial,dtype=tf.float32)init_op = tf.global_variables_initializer()with tf.Session() as sess:    sess.run(init_op)    print sess.run(tf.reduce_mean(x))    print sess.run(tf.reduce_mean(x,0)) #Column    print sess.run(tf.reduce_mean(x,1)) #row

Then execute:

python /home/ubuntu/reduce_mean.py

Execution Result:

1.5[ 1.5  1.5][ 1.  2.]

You can also try to modify the source code to see what happens to the output?

Tf.squared_difference

squared_difference(    x,    y,    name=None)

Function Description:

Calculate tensor x, y corresponding element squared difference

Parameter list:

Name of parameter	must-Choose	type	Description
X	Is	Tensor	Is half, float32, Float64, Int32, Int64, Complex64, complex128 one of the types
Y	Is	Tensor	Is half, float32, Float64, Int32, Int64, Complex64, complex128 one of the types
Name	Whether	String	Operation name

Example code:

You can now create the source file squared_difference.pyin the /home/ubuntu directory, which can be referenced in the following:

Example code:/home/ubuntu/squared_difference.py

#!/usr/bin/pythonimport tensorflow as tfimport numpy as npinitial_x = [[1.,1.],[2.,2.]]x = tf.Variable(initial_x,dtype=tf.float32)initial_y = [[3.,3.],[4.,4.]]y = tf.Variable(initial_y,dtype=tf.float32)diff = tf.squared_difference(x,y)init_op = tf.global_variables_initializer()with tf.Session() as sess:    sess.run(init_op)    print sess.run(diff)

Then execute:

python /home/ubuntu/squared_difference.py

Execution Result:

[[ 4.  4.] [ 4.  4.]]

You can also try to modify the source code to see what happens to the output?

Tf.square

square(    x,    name=None)

Function Description:

Calculates the square of the tensor corresponding element

Parameter list:

Name of parameter	must-Choose	type	Description
X	Is	Tensor	Is half, float32, Float64, Int32, Int64, Complex64, complex128 one of the types
Name	Whether	String	Operation name

Example code:

You can now create the source file square.pyin the /home/ubuntu directory, which can be referenced in the following:

Example code:/home/ubuntu/square.py

#!/usr/bin/pythonimport tensorflow as tfimport numpy as npinitial_x = [[1.,1.],[2.,2.]]x = tf.Variable(initial_x,dtype=tf.float32)x2 = tf.square(x)init_op = tf.global_variables_initializer()with tf.Session() as sess:    sess.run(init_op)    print sess.run(x2)

Then execute:

python /home/ubuntu/square.py

Execution Result:

[[ 1.  1.] [ 4.  4.]]

You can also try to modify the source code to see what happens to the output?

TensorFlow Related class understanding

Task Time: Unknown time

Tf. Variable

__init__(    initial_value=None,    trainable=True,    collections=None,    validate_shape=True,    caching_device=None,    name=None,    variable_def=None,    dtype=None,    expected_shape=None,    import_scope=None)

Function Description:

Maintain the state information during the execution of the diagram, such as changes in the weight values of the neural network.

Parameter list:

Name of parameter	type	Description
Initial_value	Tensor	The initial value of the Variable class, which must specify the shape information, or the subsequent validate_shape should be set to False
Trainable	Boolean	Whether to add variables to collection Graphkeys.trainable_variables (collection is a global store that is not affected by the variable name living space, where it is saved, and is desirable everywhere)
Collections	Graph Collections	Global storage, default is Graphkeys.global_variables
Validate_shape	Boolean	Whether to allow Initial_value initialization by an unknown dimension
Caching_device	String	Indicates which device is used to cache variables
Name	String	Variable name
Dtype	Dtype	If it is set, the initialized value will initialize as this type.
Expected_shape	Tensorshape	If set, then the initial value will be this dimension

Example code:

You can now create the source file variable.pyin the /home/ubuntu directory, which can be referenced in the following:

Example code:/home/ubuntu/variable.py

 #!/usr/bin/pythonimport TensorFlow as Tfinitial = Tf.truncated_normal (shape=[10,10],mean=0,stddev=1) W=tf. Variable (initial) list = [[1.,1.],[2.,2.] X = tf. Variable (list,dtype=tf.float32) init_op = Tf.global_variables_initializer () with TF. Session () as Sess:sess.run (init_op) print ("################## (1) ################") Print Sess.run (W) print ( "################## (2) ################") Print Sess.run (W[:2,:2]) op = w[:2,:2].assign (22.*tf.ones ((2,2))) print ("################### (3) ###############") Print Sess.run (OP) print ("################### (4) ###############") Prin T (W.eval (Sess)) #computes and returns the value of this variable print ("#################### (5) ##############") pr Int (W.eval ()) #Usage with the default session print ("##################### (6) #############") Print W.dtype pri NT Sess.run (w.initial_value) print Sess.run (w.op) Print w.shape print ("################### (7) ###############") Print Sess.run (X) 

Then execute:

python /home/ubuntu/Variable.py

Complete

Task Time: Unknown time

Congratulations, you have completed the contents of this experiment

You can do a series of more TensorFlow tutorials:

• TensorFlow-Linear regression
• TensorFlow-based on CNN digital recognition

For more information about TensorFlow, refer to TensorFlow website .

tensorflow-related APIs