In [1]:
import numpy as np
import numpy.random as rng
import tensorflow as tf
In [2]:
din = rng.randn(100,1)
out = din * .2 + .1+ .1 * rng.normal(size=din.shape)
In [3]:
%matplotlib inline
import matplotlib.pyplot as plt
In [4]:
plt.scatter(din, out)
Out[4]:
<matplotlib.collections.PathCollection at 0x10d38ad30>
In [52]:
x = tf.placeholder(np.float32, [None])
w = tf.Variable([0.0])
b = tf.Variable([0.0])
y = tf.mul(x, w) + b
y_ = tf.placeholder("float")
loss = tf.reduce_sum(tf.pow(y_ - y, 2))
init = tf.initialize_all_variables()
optim = tf.train.GradientDescentOptimizer(.005).minimize(loss)
with tf.Session() as sess:
feed_dict={x: din, y_: out}
sess.run(init)
print(sess.run(loss, feed_dict=feed_dict))
for step in range(50):
sess.run(optim, feed_dict=feed_dict)
cost_curr = sess.run(loss, feed_dict=feed_dict)
print("Iteration", step, "Loss", cost_curr)
print(sess.run(w))
beta = [sess.run(w), sess.run(b)]
6.5261 Iteration 0 Loss 0.877566 [ 0.21694028] Iteration 1 Loss 0.834038 [ 0.22164701] Iteration 2 Loss 0.833522 [ 0.22377081] Iteration 3 Loss 0.833514 [ 0.2239479] Iteration 4 Loss 0.833514 [ 0.22397718] Iteration 5 Loss 0.833514 [ 0.22398056] Iteration 6 Loss 0.833515 [ 0.22398102] Iteration 7 Loss 0.833514 [ 0.22398108] Iteration 8 Loss 0.833514 [ 0.22398108] Iteration 9 Loss 0.833514 [ 0.22398108] Iteration 10 Loss 0.833514 [ 0.22398108] Iteration 11 Loss 0.833514 [ 0.22398108] Iteration 12 Loss 0.833514 [ 0.22398108] Iteration 13 Loss 0.833514 [ 0.22398108] Iteration 14 Loss 0.833514 [ 0.22398108] Iteration 15 Loss 0.833514 [ 0.22398108] Iteration 16 Loss 0.833514 [ 0.22398108] Iteration 17 Loss 0.833514 [ 0.22398108] Iteration 18 Loss 0.833514 [ 0.22398108] Iteration 19 Loss 0.833514 [ 0.22398108] Iteration 20 Loss 0.833514 [ 0.22398108] Iteration 21 Loss 0.833514 [ 0.22398108] Iteration 22 Loss 0.833514 [ 0.22398108] Iteration 23 Loss 0.833514 [ 0.22398108] Iteration 24 Loss 0.833514 [ 0.22398108] Iteration 25 Loss 0.833514 [ 0.22398108] Iteration 26 Loss 0.833514 [ 0.22398108] Iteration 27 Loss 0.833514 [ 0.22398108] Iteration 28 Loss 0.833514 [ 0.22398108] Iteration 29 Loss 0.833514 [ 0.22398108] Iteration 30 Loss 0.833514 [ 0.22398108] Iteration 31 Loss 0.833514 [ 0.22398108] Iteration 32 Loss 0.833514 [ 0.22398108] Iteration 33 Loss 0.833514 [ 0.22398108] Iteration 34 Loss 0.833514 [ 0.22398108] Iteration 35 Loss 0.833514 [ 0.22398108] Iteration 36 Loss 0.833514 [ 0.22398108] Iteration 37 Loss 0.833514 [ 0.22398108] Iteration 38 Loss 0.833514 [ 0.22398108] Iteration 39 Loss 0.833514 [ 0.22398108] Iteration 40 Loss 0.833514 [ 0.22398108] Iteration 41 Loss 0.833514 [ 0.22398108] Iteration 42 Loss 0.833514 [ 0.22398108] Iteration 43 Loss 0.833514 [ 0.22398108] Iteration 44 Loss 0.833514 [ 0.22398108] Iteration 45 Loss 0.833514 [ 0.22398108] Iteration 46 Loss 0.833514 [ 0.22398108] Iteration 47 Loss 0.833514 [ 0.22398108] Iteration 48 Loss 0.833514 [ 0.22398108] Iteration 49 Loss 0.833514 [ 0.22398108]
In [53]:
plt.scatter(din, out)
x = np.linspace(din.min(), din.max(), 101)
plt.plot(x, beta[0] * x + beta[1])
Out[53]:
[<matplotlib.lines.Line2D at 0x1118249e8>]
In [54]:
np.polyfit(din, out ,1)
Out[54]:
array([ 0.22398109, 0.08161766])
In [ ]: