Source that deep learning with mnist data

MNIST reads the data one-hot to create tansorboard data and stores the training results.

%autosave 0

def reset_graph(seed=42):
    tf.reset_default_graph()
    tf.set_random_seed(seed)
    np.random.seed(seed)

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)

import tensorflow as tf
import numpy as np

n_inputs = 28*28  # MNIST
n_hidden1 = 300
n_hidden2 = 100
n_hidden3 = 150
n_outputs = 10
logs_path = 'tensorboard_logs'

reset_graph()
X = tf.placeholder(tf.float32, shape=(None, n_inputs), name="X")
y = tf.placeholder(tf.int64, shape=(None, n_outputs), name="y")

with tf.name_scope("dnn"):
    hidden1 = tf.layers.dense(inputs=X, units=n_hidden1, activation=tf.nn.relu)
    hidden2 = tf.layers.dense(inputs=hidden1, units=n_hidden2, activation=tf.nn.relu)
    hidden3 = tf.layers.dense(inputs=hidden2, units=n_hidden3, activation=tf.nn.relu)
    logits = tf.layers.dense(inputs=hidden2, units=n_outputs)

with tf.name_scope("loss"):
    xentropy = tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=logits)
    loss = tf.reduce_mean(xentropy, name="loss")


learning_rate = 0.01

with tf.name_scope("train"):
    optimizer = tf.train.GradientDescentOptimizer(learning_rate)
    training_op = optimizer.minimize(loss)

with tf.name_scope("eval"):
    correct = tf.nn.in_top_k(logits, tf.argmax(y,1), 1)
    accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))

init = tf.global_variables_initializer()
saver = tf.train.Saver()

n_epochs = 40
batch_size = 50
tf.summary.scalar("loss", loss)
tf.summary.scalar("accuracy", accuracy)
merged_summary_op = tf.summary.merge_all()

with tf.Session() as sess:
    init.run()
    count = 0
    summary_writer = tf.summary.FileWriter(logs_path, graph=tf.get_default_graph())
    for epoch in range(n_epochs):
        for iteration in range(mnist.train.num_examples // batch_size):
            X_batch, y_batch = mnist.train.next_batch(batch_size)
            _, summary = sess.run([training_op, merged_summary_op], feed_dict={X: X_batch, y: y_batch})
            count += 1
            summary_writer.add_summary(summary, global_step=count)
        acc_train = accuracy.eval(feed_dict={X: X_batch, y: y_batch})
        acc_test = accuracy.eval(feed_dict={X: mnist.test.images, y: mnist.test.labels})
        print(epoch, "Train accuracy:", acc_train, "Test accuracy:", acc_test)
 
    tf.add_to_collection("X", X)
    tf.add_to_collection("logits", logits)
    save_path = saver.save(sess, "./my_model_final.ckpt")

Reads the trained model and produces results.

import tensorflow as tf
import numpy as np

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)

tf.reset_default_graph()
init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)
    saver = tf.train.import_meta_graph("./my_model_final.ckpt.meta")
    saver.restore(sess, "./my_model_final.ckpt") 
    logits = tf.get_collection("logits")[0]
    X = tf.get_collection("X")[0]
    X_new_scaled = mnist.test.images[:20]
    Z = logits.eval(feed_dict={X: X_new_scaled})
    y_pred = np.argmax(Z, axis=1)
    
print("Predicted classes:", y_pred)
print("Actual classes:   ", mnist.test.labels[:20])