The simplest least squares with TensorFlow

least_squares_1.py

import tensorflow as tf

x = tf.placeholder(tf.float32, [None, 1])
y_ = tf.placeholder(tf.float32, [None, 1])

b = tf.Variable(tf.zeros([1]))
w = tf.Variable(tf.zeros([1, 1]))

y = w * x + b

loss = tf.reduce_sum((y - y_) * (y - y_))
train_step = tf.train.GradientDescentOptimizer(0.005).minimize(loss)

sess = tf.Session()
sess.run(tf.initialize_all_variables())

for step in range(10):
    sess.run(train_step, feed_dict={x:[[2.3],[1.7],[-3.8],[0.5],[-4.1],[-1.5],[-2.5],[6.2]],
                                    y_:[[-4.4],[-3.6],[7.7],[-0.9],[8.3],[2.9],[4.9],[-12.2]]})
    print step, sess.run(w), sess.run(b)

Rewrite as 'print(step, sess.run(w), sess.run(b))'.
But this code was written many years ago. Other lines do not work for the current version of TensorFlow. Please google similar codes.