import tensorflow as tf
import numpy as np
xs = np.random.randint(46, 99, 100)
ys=1.7*xs
x=tf.placeholder(tf.float32)
y=tf.placeholder(tf.float32)
w=tf.Variable(0.1)
b=tf.Variable(0.1)
y_ = tf.mul(w,x)+b  #tf.multiply
cost=tf.reduce_sum(tf.pow(y-y_,2))
train_step=tf.train.GradientDescentOptimizer(0.02).minimize(cost)
init=tf.initialize_all_variables()
sess=tf.Session()
sess.run(init)
for _ in range(10):
    sess.run(train_step,feed_dict={x:xs,y:ys})