Introduction to ML

December 13, 2018

인프런 머신러닝과 딥 러닝 강좌를 듣고 정리 한 것입니다.

Tensorflow 기본적인 operations

pip3 install --upgrade tensorflow
# or
pip3 install --upgrade tensorflow-gpu

import tensorflow as tf

tf.__version__

hello = tf.constant("Hello World!")

sess = tf.Session()

print(sess.run(hello))

n1 = tf.constant(3.0, tf.float32)

n2 = tf.constant(4.0)

n3 = tf.add(n1, n2)

sess.run([n1, n2])

sess.run(n3)

Placeholder

a = tf.placeholder(tf.float32)

b = tf.placeholder(tf.float32)

adder_node = a + b

sess.run(adder_node, feed_dict={a: 3, b: 4.5}))
# 7.5

sess.run(adder_node, feed_dict={a: [1,3], b: [2, 4]}))
# [ 3. 7.]

Type, Shape, Rank

rank

shape

type

examples

Linear Regression

Hypothesis based on “H(x)=Wx+b”

Cost: (H(x)-y)^2

total_cost

Goal: Minimize “cost(W,b)”

Linear Regression의 cost 최소화 알고리즘 원리

x = [1, 2, 3]

y = [2, 4, 6]

W = tf.Variable(tf.random_normal([1]), name='weight')

b = tf.Variable(tf.random_normal([1]), name='bias')

hypo = x*W+b

cost=tf.reduce_mean(tf.square(hypo-y))

optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.01)

train = optimizer.minimize(cost)

sess = tf.Session()

sess.run(tf.global_variables_initializer())

for step in range(2001):
	sess.run(train)
	if step % 20 == 0:
		print(step, sess.run(cost), sess.run(W), sess.run(b))

Placeholder로 해보기

X = tf.placeholder(tf.float32)
Y = tf.placeholder(tf.float32)

for step in range(2001):
	cost_val, W_val, b_val, _ = \
		sess.run([cost, W, b, train],
			feed_dict={X: [1, 2, 3, 4, 5], Y: [2.1, 3.1, 4.1, 5.1, 6.1]})
	if step % 20 == 0:
		print(step, cost_val, W_val, b_val)