TF/05_Nearest_Neighbor_Methods/03 Working with Text Distances04 Computing with Mixed Distance Functi

03 Working with Text Distances

# Text Distances
#----------------------------------
#
# This function illustrates how to use
# the Levenstein distance (edit distance)
# in TensorFlow.

import tensorflow as tf

sess = tf.Session()

#----------------------------------
# First compute the edit distance between 'bear' and 'beers'
hypothesis = list('bear')
truth = list('beers')
h1 = tf.SparseTensor([[0,0,0], [0,0,1], [0,0,2], [0,0,3]],
                     hypothesis,
                     [1,1,1])

t1 = tf.SparseTensor([[0,0,0], [0,0,1], [0,0,2], [0,0,3],[0,0,4]],
                     truth,
                     [1,1,1])

print(sess.run(tf.edit_distance(h1, t1, normalize=False)))

#----------------------------------
# Compute the edit distance between ('bear','beer') and 'beers':
hypothesis2 = list('bearbeer')
truth2 = list('beersbeers')
h2 = tf.SparseTensor([[0,0,0], [0,0,1], [0,0,2], [0,0,3], [0,1,0], [0,1,1], [0,1,2], [0,1,3]],
                     hypothesis2,
                     [1,2,4])

t2 = tf.SparseTensor([[0,0,0], [0,0,1], [0,0,2], [0,0,3], [0,0,4], [0,1,0], [0,1,1], [0,1,2], [0,1,3], [0,1,4]],
                     truth2,
                     [1,2,5])

print(sess.run(tf.edit_distance(h2, t2, normalize=True)))

#----------------------------------
# Now compute distance between four words and 'beers' more efficiently:
hypothesis_words = ['bear','bar','tensor','flow']
truth_word = ['beers']

num_h_words = len(hypothesis_words)
h_indices = [[xi, 0, yi] for xi,x in enumerate(hypothesis_words) for yi,y in enumerate(x)]
h_chars = list(''.join(hypothesis_words))

h3 = tf.SparseTensor(h_indices, h_chars, [num_h_words,1,1])

truth_word_vec = truth_word*num_h_words
t_indices = [[xi, 0, yi] for xi,x in enumerate(truth_word_vec) for yi,y in enumerate(x)]
t_chars = list(''.join(truth_word_vec))

t3 = tf.SparseTensor(t_indices, t_chars, [num_h_words,1,1])

print(sess.run(tf.edit_distance(h3, t3, normalize=True)))

[[ 2.]]
[[ 0.40000001  0.2       ]]
[[ 0.40000001]
 [ 0.60000002]
 [ 1.        ]
 [ 1.        ]]

04 Computing with Mixed Distance Functions

有问题啊
04_mixed_distance_functions_knn.py

# Mixed Distance Functions for  k-Nearest Neighbor
#----------------------------------
#
# This function shows how to use different distance
# metrics on different features for kNN.
#
# Data:
#----------x-values-----------
# CRIM   : per capita crime rate by town
# ZN     : prop. of res. land zones
# INDUS  : prop. of non-retail business acres
# CHAS   : Charles river dummy variable
# NOX    : nitrix oxides concentration / 10 M
# RM     : Avg. # of rooms per building
# AGE    : prop. of buildings built prior to 1940
# DIS    : Weighted distances to employment centers
# RAD    : Index of radian highway access
# TAX    : Full tax rate value per $10k
# PTRATIO: Pupil/Teacher ratio by town
# B      : 1000*(Bk-0.63)^2, Bk=prop. of blacks
# LSTAT  : % lower status of pop
#------------y-value-----------
# MEDV   : Median Value of homes in $1,000's


import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
import requests
from tensorflow.python.framework import ops
ops.reset_default_graph()

# Create graph
sess = tf.Session()

# Load the data
housing_url = 'https://archive.ics.uci.edu/ml/machine-learning-databases/housing/housing.data'
housing_header = ['CRIM', 'ZN', 'INDUS', 'CHAS', 'NOX', 'RM', 'AGE', 'DIS', 'RAD', 'TAX', 'PTRATIO', 'B', 'LSTAT', 'MEDV']
cols_used = ['CRIM', 'INDUS', 'NOX', 'RM', 'AGE', 'DIS', 'TAX', 'PTRATIO', 'B', 'LSTAT']
num_features = len(cols_used)
housing_file = requests.get(housing_url)
housing_data = [[float(x) for x in y.split(' ') if len(x)>=1] for y in housing_file.text.split('\n') if len(y)>=1]

y_vals = np.transpose([np.array([y[13] for y in housing_data])])
x_vals = np.array([[x for i,x in enumerate(y) if housing_header[i] in cols_used] for y in housing_data])

## Min-Max Scaling
x_vals = (x_vals - x_vals.min(0)) / x_vals.ptp(0)

## Create distance metric weight matrix weighted by standard deviation
weight_diagonal = x_vals.std(0)
weight_matrix = tf.cast(tf.diag(weight_diagonal), dtype=tf.float32)

# Split the data into train and test sets
np.random.seed(13)   # reproducible results
train_indices = np.random.choice(len(x_vals), round(len(x_vals)*0.8), replace=False)
test_indices = np.array(list(set(range(len(x_vals))) - set(train_indices)))
x_vals_train = x_vals[train_indices]
x_vals_test = x_vals[test_indices]
y_vals_train = y_vals[train_indices]
y_vals_test = y_vals[test_indices]

# Declare k-value and batch size
k = 4
batch_size=len(x_vals_test)

# Placeholders
x_data_train = tf.placeholder(shape=[None, num_features], dtype=tf.float32)
x_data_test = tf.placeholder(shape=[None, num_features], dtype=tf.float32)
y_target_train = tf.placeholder(shape=[None, 1], dtype=tf.float32)
y_target_test = tf.placeholder(shape=[None, 1], dtype=tf.float32)

# Declare weighted distance metric
# Weighted - L2 = sqrt((x-y)^T * A * (x-y))
subtraction_term =  tf.subtract(x_data_train, tf.expand_dims(x_data_test,1))
first_product = tf.matmul(subtraction_term, tf.tile(tf.expand_dims(weight_matrix,0), [batch_size,1,1]))
second_product = tf.matmul(first_product, tf.transpose(subtraction_term, perm=[0,2,1]))
distance = tf.sqrt(tf.matrix_diag_part(second_product))

# Predict: Get min distance index (Nearest neighbor)
top_k_xvals, top_k_indices = tf.nn.top_k(tf.negative(distance), k=k)
x_sums = tf.expand_dims(tf.reduce_sum(top_k_xvals, 1),1)
x_sums_repeated = tf.matmul(x_sums,tf.ones([1, k], tf.float32))
x_val_weights = tf.expand_dims(tf.div(top_k_xvals,x_sums_repeated), 1)

top_k_yvals = tf.gather(y_target_train, top_k_indices)
prediction = tf.squeeze(tf.matmul(x_val_weights,top_k_yvals), axis=[1])

# Calculate MSE
mse = tf.div(tf.reduce_sum(tf.square(tf.subtract(prediction, y_target_test))), batch_size)

# Calculate how many loops over training data
num_loops = int(np.ceil(len(x_vals_test)/batch_size))

for i in range(num_loops):
    min_index = i*batch_size
    max_index = min((i+1)*batch_size,len(x_vals_train))
    x_batch = x_vals_test[min_index:max_index]
    y_batch = y_vals_test[min_index:max_index]
    predictions = sess.run(prediction, feed_dict={x_data_train: x_vals_train, x_data_test: x_batch,
                                         y_target_train: y_vals_train, y_target_test: y_batch})
    batch_mse = sess.run(mse, feed_dict={x_data_train: x_vals_train, x_data_test: x_batch,
                                         y_target_train: y_vals_train, y_target_test: y_batch})

    print('Batch #' + str(i+1) + ' MSE: ' + str(np.round(batch_mse,3)))

# Plot prediction and actual distribution
bins = np.linspace(5, 50, 45)

plt.hist(predictions, bins, alpha=0.5, label='Prediction')
plt.hist(y_batch, bins, alpha=0.5, label='Actual')
plt.title('Histogram of Predicted and Actual Values')
plt.xlabel('Med Home Value in $1,000s')
plt.ylabel('Frequency')
plt.legend(loc='upper right')
plt.show()