In [1]:

    

import numpy as np
import tensorflow as tf
import iris
import matplotlib.pyplot as plt
from neural_network_decision_tree import nn_decision_tree
%matplotlib inline


    

In [2]:

    

tf.__version__


    

    
Out[2]:





'1.9.0'

In [3]:

    

np.random.seed(1943)
tf.set_random_seed(1943)


    

In [4]:

    

x = iris.feature[:, 2:4]  # use "Petal length" and "Petal width" only
y = iris.label
d = x.shape[1]


    

In [5]:

    

num_cut = [1, 1]  # "Petal length" and "Petal width"
num_leaf = np.prod(np.array(num_cut) + 1)
num_class = 3


    

In [6]:

    

sess = tf.InteractiveSession()


    

In [7]:

    

x_ph = tf.placeholder(tf.float32, [None, d])
y_ph = tf.placeholder(tf.float32, [None, num_class])


    

In [8]:

    

cut_points_list = [tf.Variable(tf.random_uniform([i])) for i in num_cut]
leaf_score = tf.Variable(tf.random_uniform([num_leaf, num_class]))


    

In [9]:

    

y_pred = nn_decision_tree(x_ph, cut_points_list, leaf_score, temperature=0.1)
loss = tf.reduce_mean(tf.losses.softmax_cross_entropy(logits=y_pred, onehot_labels=y_ph))


    

In [10]:

    

opt = tf.train.AdamOptimizer(0.1)
train_step = opt.minimize(loss)


    

In [11]:

    

sess.run(tf.global_variables_initializer())


    

In [12]:

    

for i in range(1000):
    _, loss_e = sess.run([train_step, loss], feed_dict={x_ph: x, y_ph: y})
    if i % 200 == 0:
        print(loss_e)
print('error rate %.2f' % (1 - np.mean(np.argmax(y_pred.eval(feed_dict={x_ph: x}), axis=1) == np.argmax(y, axis=1))))


    

    




1.1931068
0.46273762
0.45842612
0.45060724
0.118448496
error rate 0.04

In [13]:

    

sample_x0 = np.repeat(np.linspace(0, np.max(x[:,0]), 100), 100).reshape(-1,1)
sample_x1 = np.tile(np.linspace(0, np.max(x[:,1]), 100).reshape(-1,1), [100,1])
sample_x = np.hstack([sample_x0, sample_x1])
sample_label = np.argmax(y_pred.eval(feed_dict={x_ph: sample_x}), axis=1)


    

In [14]:

    

plt.figure(figsize=(8,8))

plt.scatter(x[:,0], 
            x[:,1], 
            c=np.argmax(y, axis=1), 
            marker='o',
            s=50,
            cmap='summer', 
            edgecolors='black')

plt.scatter(sample_x0.flatten(), 
            sample_x1.flatten(), 
            c=sample_label.flatten(), 
            marker='D',
            s=20,
            cmap='summer', 
            edgecolors='none',
            alpha=0.33)


    

    
Out[14]:





<matplotlib.collections.PathCollection at 0x2c80669e1d0>