10.

2
In [1]:
from keras.preprocessing.text import Tokenizer

from keras.preprocessing.sequence import pad_sequences

import numpy as np

import matplotlib.pyplot as plt

from pathlib import Path

from keras.models import Sequential

from keras.layers import Embedding, Flatten, Dense

import os

from contextlib import redirect_stdout

import time

start_time = time.time()

In [8]:
results_dir = Path('results').joinpath('model_1')

results_dir.mkdir(parents=True, exist_ok=True)

# path for IMDB data

imdb_dir = Path('/home/desktop/dsc650/data/external/imdb/aclImdb/')
test_dir = os.path.join(imdb_dir, 'test')

train_dir = os.path.join(imdb_dir, 'train')

In [9]:
training_samples = 200

maxlen = 100 # cuts off review after 100 words

max_words = 1000 # Considers only the top 1000 words in the dataset
embedding_dim = 100

training_samples = 200 # trains on 200 samples

validation_samples = 10000 # validates on 10000 samples

In [10]:
labels = []

texts = []

for label_type in ['neg', 'pos']:

dir_name = os.path.join(test_dir, label_type)

for fname in sorted(os.listdir(dir_name)):

if fname[-4:] == '.txt':

f = open(os.path.join(dir_name, fname), encoding="utf8")

texts.append(f.read())

f.close()

if label_type == 'neg':

labels.append(0)

else:

labels.append(1)

In [11]:
tokenizer = Tokenizer(num_words=max_words)

tokenizer.fit_on_texts(texts)

sequences = tokenizer.texts_to_sequences(texts)

word_index = tokenizer.word_index
print('Found %s unique tokens.' % len(word_index))

data = pad_sequences(sequences, maxlen=maxlen)

labels = np.asarray(labels)

print('Shape of data tensor:', data.shape)

print('Shape of label tensor:', labels.shape)

Found 87393 unique tokens.

Shape of data tensor: (25000, 100)

Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js
Shape of label tensor: (25000,)

 In [12]:
indices = np.arange(data.shape[0])

np.random.shuffle(indices)

data = data[indices]

labels = labels[indices]

x_train = data[:training_samples]
y_train = labels[:training_samples]

x_val = data[training_samples: training_samples + validation_samples]

y_val = labels[training_samples: training_samples + validation_samples]

In [13]:
# from page 191 Listing 6.12

model = Sequential()

model.add(Embedding(max_words, embedding_dim, input_length=maxlen))

model.add(Flatten())

model.add(Dense(32,activation='relu'))

model.add(Dense(1, activation='sigmoid'))

In [14]:
# Save the summary to file

summary_file = results_dir.joinpath('Assignment_10.2_ModelSummary.txt')

with open(summary_file, 'w') as f:

with redirect_stdout(f):

model.summary()

# from page 192

model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])

history=model.fit(x_train, y_train, epochs=10, batch_size=32,validation_data=(x_val,

result_model_file = results_dir.joinpath('pre_trained_glove_model.h5')

model.save_weights(result_model_file)

Epoch 1/10

7/7 [==============================] - 1s 120ms/step - loss: 0.6966 - acc: 0.4703 -

val_loss: 0.6928 - val_acc: 0.5115
Epoch 2/10

7/7 [==============================] - 1s 101ms/step - loss: 0.5752 - acc: 0.9481 -

val_loss: 0.6933 - val_acc: 0.5143
Epoch 3/10

7/7 [==============================] - 1s 96ms/step - loss: 0.4203 - acc: 0.9792 - v

al_loss: 0.6991 - val_acc: 0.5142

Epoch 4/10

7/7 [==============================] - 1s 104ms/step - loss: 0.2528 - acc: 0.9674 -

val_loss: 0.7005 - val_acc: 0.5265
Epoch 5/10

7/7 [==============================] - 1s 99ms/step - loss: 0.1495 - acc: 0.9931 - v

al_loss: 0.7039 - val_acc: 0.5273

Epoch 6/10

7/7 [==============================] - 1s 99ms/step - loss: 0.0822 - acc: 0.9963 - v

al_loss: 0.7383 - val_acc: 0.5208

Epoch 7/10

7/7 [==============================] - 1s 100ms/step - loss: 0.0541 - acc: 1.0000 -

val_loss: 0.7206 - val_acc: 0.5307
Epoch 8/10

7/7 [==============================] - 1s 99ms/step - loss: 0.0295 - acc: 1.0000 - v

al_loss: 0.7405 - val_acc: 0.5246

Epoch 9/10

7/7 [==============================] - 1s 101ms/step - loss: 0.0179 - acc: 1.0000 -

val_loss: 0.7399 - val_acc: 0.5295
Epoch 10/10

7/7 [==============================] - 1s 101ms/step - loss: 0.0099 - acc: 1.0000 -

val_loss: 0.7519 - val_acc: 0.5313

In [15]:
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js
# Plots

acc = history.history['acc']

 val_acc = history.history['val_acc']

loss = history.history['loss']

val_loss = history.history['val_loss']

epochs = range(1, len(acc) + 1)

plt.plot(epochs, acc, 'bo', label='Training acc')

plt.plot(epochs, val_acc, 'b', label='Validation acc')

plt.title('Training and validation accuracy')

plt.legend()

plt.figure()

plt.plot(epochs, loss, 'bo', label='Training loss')

plt.plot(epochs, val_loss, 'b', label='Validation loss')

plt.title('Training and validation loss')

plt.legend()

img_file = results_dir.joinpath('Assignment_10.2_Model Accuracy Validation.png')

plt.savefig(img_file)

plt.show()

In [16]:
labels=[]

texts=[]

for label_type in ['neg', 'pos']:

dir_name = os.path.join(test_dir, label_type)

for fname in sorted(os.listdir(dir_name)):

if fname[-4:] == '.txt':

f = open(os.path.join(dir_name, fname), encoding="utf8")

texts.append(f.read())

f.close()

if label_type == 'neg':

labels.append(0)

Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js
else:

 labels.append(1)

In [17]:
sequence = tokenizer.texts_to_sequences(texts)

x_test = pad_sequences(sequences, maxlen=maxlen)

y_test = np.asarray(labels)

model.load_weights(result_model_file)

eval = model.evaluate(x_test, y_test)

print("")

print(eval)

print("Complete: --- %s seconds has passed ---" % (time.time() - start_time))

782/782 [==============================] - 2s 2ms/step - loss: 0.7500 - acc: 0.5329

[0.750030517578125, 0.5329200029373169]

Complete: --- 453.8152759075165 seconds has passed ---

In [ ]:

Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js