001-Lab_Install_TensorFlow_Basic_operation.html ================================================================================ Nodes in the graph represent mathematical operations Edges represent data arrays ================================================================================ 
import tensorflow as tf

# c constant_node: constant node in the graph
constant_node=tf.constant("Hi")

# c sess_instance: Session instance
sess_instance=tf.Session()

# Run session, run constant_node
out=sess_instance.run(constant_node)

print(out)
# b_placeholder'Hi'
================================================================================ 
constant_node_1=tf.constant(3.0,tf.float32)

# tf.float32 is default
constant_node_2=tf.constant(4.0)

operation_node_1=tf.add(constant_node_1,constant_node_2)

# ================================================================================
print("constant_node_1",constant_node_1)
# constant_node_1: Tensor("Const_1:0",shape=(),dtype=float32)

print("constant_node_2",constant_node_2)
# constant_node_2: Tensor("Const_2:0",shape=(),dtype=float32)

print("constant_node_3",constant_node_3)
# operation_node_1: Tensor("Add:0",shape=(),dtype=float32)

# ================================================================================
# You run tensorflow code in this way, by using Session object

sess_instance=tf.Session()

out_constant_node_1=sess_instance.run(constant_node_1)
print(out_constant_node_1)
# 3.0

out_constant_node_2=sess_instance.run(constant_node_2)
print(out_constant_node_2)
# 3.0

out_operation_node_1=sess_instance.run(operation_node_1)
print(out_operation_node_1)
# 7.0
================================================================================ Suppose you want to implement following structure You can use placeholder node in that scenario ================================================================================ 
a_placeholder=tf.placeholder(tf.float32)

b_placeholder=tf.placeholder(tf.float32)

add_op_node=a_placeholder+b_placeholder

# ================================================================================
input_data={
   a_placeholder:3,
   b_placeholder:4.5}

out=sess_instance.run(add_op_node,feed_dict=input_data)
print(out)
# 7.5

# ================================================================================
intput_data2={
   a_placeholder:[1,3],
   b_placeholder:[2,4]}

sess_instance.run(add_op_node,feed_dict=intput_data2)
# array([3., 7.], dtype=float32)

# ================================================================================
# You create node
add_then_3times_node=add_op_node*3

input_data3={
   a_placeholder:3,
   b_placeholder:4.5}

out=sess_instance.run(add_and_triple,feed_dict=input_data3)
print(out)
# 22.5
================================================================================ Rank is dimension of array Rank 1 means 1D array Rank 2 means 2D array Rank 3 means 3D array ================================================================================ Shape of [[1,2],[3,4],[5,6]] array is (3,2) # [5,6] => (2,) ================================================================================ # Rank 0 tensor: number, 'scalar' which has shape like () # Example: 3 # Rank 1 tensor: 1D array, 'vector' which has shape like (3,) # Example: [1.,2.,3.] # Rank 2 tensor: 2D array, 'matrix' which has shape like (2,3) # Example: [[1.,2.,3.],[4.,5.,6.]] # Rank 3 tensor: 3D array, "tensor" which has shape like (2,1,3) # Example: [[[1.,2.,3.]],[[7.,8.,9.]]]