How to define TensorFlow input nodes

There are several ways to define input nodes in TensorFlow as follows.

Defined via placeholders: this is generally used.

Defined by dictionary type: generally used when there are more inputs.

Direct definition: generally rarely used.

I Placeholder definitions

Example:

Specifically using the function, the code is as follows:

X = ("float")
Y = ("float")

II Dictionary Type Definitions

1 Example

Defining input nodes by dictionary type

2 Key Code

# Create models
# placeholders
inputdict = {
  'x': ("float"),
  'y': ("float")
}

3 Interpretation

Defining by dictionary is more like the first one, except that it is stacked together.

4 All codes

import tensorflow as tf
import numpy as np
import  as plt
plotdata = { "batchsize":[], "loss":[] }
def moving_average(a, w=10):
  if len(a) < w:
    return a[:]  
  return [val if idx < w else sum(a[(idx-w):idx])/w for idx, val in enumerate(a)]
# Generate simulated data
train_X = (-1, 1, 100)
train_Y = 2 * train_X + (*train_X.shape) * 0.3 # y=2x, but add noise
# Graphic display
(train_X, train_Y, 'ro', label='Original data')
()
()
# Create models
# placeholders
inputdict = {
  'x': ("float"),
  'y': ("float")
}
# Model parameters
W = (tf.random_normal([1]), name="weight")
b = (([1]), name="bias")
# Forward structure
z = (inputdict['x'], W)+ b
#Reverse Optimization
cost =tf.reduce_mean( (inputdict['y'] - z))
learning_rate = 0.01
optimizer = (learning_rate).minimize(cost) #Gradient descent
# Initialize variables
init = tf.global_variables_initializer()
# Parameter settings
training_epochs = 20
display_step = 2
# Start the session
with () as sess:
  (init)
  # Fit all training data
  for epoch in range(training_epochs):
    for (x, y) in zip(train_X, train_Y):
      (optimizer, feed_dict={inputdict['x']: x, inputdict['y']: y})
    # Show details of training in progress
    if epoch % display_step == 0:
      loss = (cost, feed_dict={inputdict['x']: train_X, inputdict['y']:train_Y})
      print ("Epoch:", epoch+1, "cost=", loss,"W=", (W), "b=", (b))
      if not (loss == "NA" ):
        plotdata["batchsize"].append(epoch)
        plotdata["loss"].append(loss)
  print (" Finished!")
  print ("cost=", (cost, feed_dict={inputdict['x']: train_X, inputdict['y']: train_Y}), "W=", (W), "b=", (b))
  # Graphic display
  (train_X, train_Y, 'ro', label='Original data')
  (train_X, (W) * train_X + (b), label='Fitted line')
  ()
  ()
  
  plotdata["avgloss"] = moving_average(plotdata["loss"])
  (1)
  (211)
  (plotdata["batchsize"], plotdata["avgloss"], 'b--')
  ('Minibatch number')
  ('Loss')
  ('Minibatch run vs. Training loss')
   
  ()
  print ("x=0.2，z=", (z, feed_dict={inputdict['x']: 0.2}))

5 Operational results

III. Direct definitions

1 Example

Direct definition of input results

2 Interpretation

Direct definition: defined Python variables are put directly into the OP node to participate in the input, and variables from simulated data are put directly into the model for training.

3 Code

import tensorflow as tf
import numpy as np
import  as plt
# Generate simulated data
train_X =np.float32( (-1, 1, 100))
train_Y = 2 * train_X + (*train_X.shape) * 0.3 # y=2x, but add noise
# Graphic display
(train_X, train_Y, 'ro', label='Original data')
()
()
# Create models
# Model parameters
W = (tf.random_normal([1]), name="weight")
b = (([1]), name="bias")
# Forward structure
z = (W, train_X)+ b
#Reverse Optimization
cost =tf.reduce_mean( (train_Y - z))
learning_rate = 0.01
optimizer = (learning_rate).minimize(cost) #Gradient descent
# Initialize variables
init = tf.global_variables_initializer()
# Parameter settings
training_epochs = 20
display_step = 2
# Start the session
with () as sess:
  (init)
  # Fit all training data
  for epoch in range(training_epochs):
    for (x, y) in zip(train_X, train_Y):
      (optimizer)
    # Show details of training in progress
    if epoch % display_step == 0:
      loss = (cost)
      print ("Epoch:", epoch+1, "cost=", loss,"W=", (W), "b=", (b))
  print (" Finished!")
  print ("cost=", (cost), "W=", (W), "b=", (b))

4 Operational results

The above article on how to define TensorFlow input nodes is all that I have shared with you.