loss = tf.reduce_sum(tf.square(linear_model - y)) optimizer = tf.train.GradientDescentOptimizer (0.01) train = optimizer