TensorFlow Linear Regression

TensorFlow Linear Regression

이전 포스트

TensorFlow 설치 및 PyCharm 설치

---

Lineare Regression

Hypothesis

H(x) = Wx+b ( W : weight)

Cost(Loss) Function

cost(W,b) = 1mi=1m(H(xi)-yi)2

Cost function 에서 minimize 를 하기 위해 Gradient Descent Algorithm을 사용 하였습니다.

Source

import tensorflow as tf

# Training Data

x_data = [1.,2.,3.]

y_data = [1.,2.,3.]

# Try to find values for W and b that compute y_data = W * x_data + b

# We know that W should be 1 and b 0

# random_uniform -> random function -1.0 ~ 1.0

W = tf.Variable(tf.random_uniform([1],-1.0,1.0))

b = tf.Variable(tf.random_uniform([1],-1.0,1.0))

# Place Holder

X = tf.placeholder(tf.float32)

Y = tf.placeholder(tf.float32)

# Our Hypothesis H(x) = Wx+b

hypothesis = W * X + b

# Simplified cost function Cost(Loss) = 1/m * Sigma (H(xi)-yi)2

# tf.square = x^2

cost = tf.reduce_mean(tf.square(hypothesis-Y))

# Minimize -> Gradient Descent Algorithm

a = tf.Variable(0.1) # Learning rate, alpha

optimizer = tf.train.GradientDescentOptimizer(a)

train = optimizer.minimize(cost)

# Before Starting, initalize the Variables

init = tf.initialize_all_variables()

# Launch the graph.

sess = tf.Session()

sess.run(init)

# Fit the line.

for step in xrange(2001) :

      sess.run(train, feed_dict={X:x_data, Y:y_data})

      if step % 30 == 0 :

          print step, sess.run(cost, feed_dict={X:x_data, Y:y_data}), sess.run(W), sess.run(b)

# Leanrs best fit is W : [0.1], b : [0.3]

API

tf.Variable() : 변수 생성

https://www.tensorflow.org/versions/r0.9/api_docs/python/state_ops.html#Variable

tf.random_uniform() : Random으로 실수를 뽑아낸다.

https://www.tensorflow.org/versions/r0.9/api_docs/python/constant_op.html#random_uniform

tf.placeholder() : functon 의 Parameter로 넘기지만, 함수를 실행할 때 값을 넘겨 준다.

TensorFlow provides a placeholder operation that must be fed with data on execution. For more info, see the section on Feeding data.

https://www.tensorflow.org/versions/r0.9/api_docs/python/io_ops.html#placeholders

tf.reduce_mean() : element 들의 평균을 합니다.

Computes the mean of elements across dimensions of a tensor.

https://www.tensorflow.org/versions/r0.9/api_docs/python/math_ops.html#reduce_mean

tf.square() : 제곱합니다.

Computes square of x element-wise.

I.e., y=x∗x=x2.

https://www.tensorflow.org/versions/r0.9/api_docs/python/math_ops.html#square

tf.train.GradientDescentoptimizer : Gradient descent algorithm 사용

Optimizer that implements the gradient descent algorithm.

https://www.tensorflow.org/versions/r0.9/api_docs/python/train.html#GradientDescentOptimizer

tf.initialize_all_variables() : 모든 변수 초기화

Returns an Op that initializes all variables.

https://www.tensorflow.org/versions/r0.9/api_docs/python/state_ops.html#initialize_all_variables

결과 : Step 30번을 주기로  step, cost, W, b 값을 출력합니다.

cost가 0이 되는 구간은 대략  690 step으로 W = 1 인 것을 알 수 있습니다.

이상 포스팅을 마칩니다.

---

출처

Sung Kim http://hunkim.github.io/ml/

TensorFlow https://www.tensorflow.org/

PyCharm https://www.jetbrains.com/pycharm/