File description:

load_data.py: implments the methods in lecture 5 (data preprocessing) and lecture 6 (label distribution). My implementation doesn't strictly follow tf girls' codes but has some improvements.

Notes of tf_girls

1. How does this one-hot coding np.eye(n_labels)[target_vector] works? https://stackoverflow.com/questions/45068853/how-does-this-one-hot-vector-conversion-work

2. Tuesday afternoon, 12-25-2018, finished mnist_fc_placeholder.py

   load mnist dataset

   There are two ways to load mnist dataset. One way (deprecated) is to

   from tensorflow.examples.tutorials.mnist import input_data
   
   mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
   
   test_samples, test_labels = mnist.test.next_batch(1000)
   
   mnist.test.num_examples return number of elements in test. Same for train.
   
   x_train = mnist.train.images
   y_train = mnist.train.labels
   

   A very good example can be found here https://github.com/snehalvartak/MNIST/blob/master/FullyConnectedNet.ipynb

   Another way is to load mnist from keras

   mnist = tf.keras.datasets.mnist
   
   (x_train, y_train),(x_test, y_test) = mnist.load_data()
   

3. How to visualize intermediate and final results

4. Tuesday night, 12-25-2018, finished the tutorial reading and dataset_tutorial.py

   Tutorial of "How to use Dataset in Tensorflow"

   link: https://towardsdatascience.com/how-to-use-dataset-in-tensorflow-c758ef9e4428 (In this tutorial, all the examples mix placeholder with dataset which I think totally lost the point of using dataset. Demos in cs20i are more proper examples. )

   cs20i compares placeholder and dataset in https://docs.google.com/document/d/1kMGs68rIHWHifBiqlU3j_2ZkrNj9RquGTe8tJ7eR1sE/edit

5. Wednesday, 12-26-2018, completed mnist classification using dataset rather than placeholder (mnist_fc_dataset.py)

6. Thursday, 12-27-2018, summarize the usage of dataset (learn to use jupyter notebook), figure out how shuffle works and what does buffer_size mean.

7. Friday, 12-28-2018, completed mnist classification using convNet mnist_convNet_raw_dataset.py (raw implementation). Refer to code https://github.com/aymericdamien/TensorFlow-Examples/blob/master/examples/3_NeuralNetworks/convolutional_network_raw.py

8. From 01-15-2019, practice tensorflow following examples of aymericdamien