MNISTをTensorboardに表示する

mnist_tensorboard.py

# -*- coding: utf-8 -*-

import input_data
import tensorflow as tf

# prameters
learning_rate = 0.01
training_epochs = 1000

def main():
	mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)

	# 重みと閾値
	W = tf.Variable(tf.zeros([784, 10]), name="weights")
	b = tf.Variable(tf.zeros([10]), name="bias")

	# mnist画像データ 28*28=784
	x = tf.placeholder("float", [None, 784])
	# 0-9なので10のクラス
	y = tf.placeholder("float", [None, 10])

	# softmax
	activation = tf.nn.softmax(tf.matmul(x, W) + b)

	# 損失関数
	# 正解とのズレ
	# Σ(y * log(y))
	cost = -tf.reduce_sum(y * tf.log(activation))

	# 学習の仕方を定義
	# Gradient Descent = 勾配降下法 学習率が引数
	train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)

	tf.scalar_summary("lossを表示しますよ", cost)

	# セッションを準備
	sess = tf.Session()

	# 変数を初期化
	init = tf.initialize_all_variables()
	sess.run(init)

	# ログを残す
	summary_op = tf.merge_all_summaries()
	summary_writer = tf.train.SummaryWriter("/tmp/tensorflow_logs", graph_def=sess.graph_def)

	for epoch in range(training_epochs):
		batch_xs, batch_ys = mnist.train.next_batch(100)
		# 勾配を用いた更新
		sess.run(train_step, feed_dict={x :batch_xs, y: batch_ys})

		summary_str = sess.run(summary_op,  feed_dict={x :batch_xs, y: batch_ys})
		summary_writer.add_summary(summary_str, epoch)

	print("Optimization Finished!")

	# 正答率を返す関数を定義
	correct_prediction = tf.equal(tf.argmax(activation, 1), tf.argmax(y, 1))
	# accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
	accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

	# 結果
	print(sess.run(accuracy, feed_dict={x: mnist.test.images, y: mnist.test.labels}))

if __name__ == "__main__":
    main()
    
# Run the command line: tensorboard --logdir=/tmp/tensorflow_logs