Note - Tensorflow Machine Learning Cookbook

Tensorflow work flow

1. Import or generate datasets

2. Transfer and normalize data 转换与归一化

   data=tf.nn.batch_norm_with_global_normalization(..)
   

3. Partition datasets into train,test,validation sets

   程序运行所需要的全局常量

   param learning_rate = 0.01 batch_size=100 iterations=1000

4. Set algorithm parameters(hyperparameters)

5. Initialize variables and placeholders

6. Define Model Structure

   with tf.Session(graph=graph) as session: seesion.run()

7. Declare the loss function

   #it is very important that it tell us how far off our prediction are from the actual value. loss=tf.reduce_mean(tf.square(y-acture - y-predit))

8. Initialize and train the model

9. Evaluate the model

10. Turn hyperparameters

11. Deploy/predict new outcome