logits = tf.layers.dense(hidden_layer_1, 10)
# label placeholder
label_placeholder = tf.placeholder(tf.uint8, shape=[None])
label_one_hot = tf.one_hot(label_placeholder, 10)
# loss function
loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits_v2(labels=label_one_hot,
logits=logits))
# DONE choose better backpropagation
# backpropagation algorithm
train = tf.train.AdamOptimizer().minimize(loss)
accuracy = dataUtils.accuracy(logits, label_one_hot)
# summaries
tf.summary.scalar('accuracy', accuracy)
tf.summary.scalar('loss', loss)
tf.summary.tensor_summary("logits", logits)
tf.summary.tensor_summary("labels", label_one_hot)
summary_tensor = tf.summary.merge_all()
saver = tf.train.Saver()
# Make tensorflow session
with tf.Session() as sess:
training_summary_writer = tf.summary.FileWriter(
TENSORBOARD_LOGDIR + "/training", sess.graph)
# Logit layer
logits = tf.nn.softmax(tf.layers.dense(
hidden_layer_2, 2, activation=None)) # replace with your code
# label placeholder
label_placeholder = tf.placeholder(tf.float32,
shape=[None, 2]) # replace with your code
# loss function
loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits_v2(labels=label_placeholder,
logits=logits))
# backpropagation algorithm
train = tf.train.AdamOptimizer().minimize(loss) # replace with your code
accuracy = dataUtils.accuracy(logits, label_placeholder)
# summaries
tf.summary.scalar('accuracy', accuracy)
tf.summary.scalar('loss', loss)
merged = tf.summary.merge_all()
# Add ops to save and restore all the variables.
saver = tf.train.Saver()
# Make tensorflow session
with tf.Session() as sess:
train_writer = tf.summary.FileWriter("summary/project1", sess.graph)
# Initialize variables
sess.run(tf.global_variables_initializer())
