log_writer = LogWriter(log_file_desc='{}{}'.format(test_name, ""),
result_desc="NeuralTopicModel")
log_writer.write_any('model', autoencoder.to_json(), 'w+', True)
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'g', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss {}'.format(
dataset_helper.get_dataset_name()))
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.savefig(
log_writer.get_plot_path(dataset_helper.get_dataset_name(), "loss"))
plt.clf()
"""topic_words_in = [sorted(topic_words,key=lambda x: x[1],reverse=True) for topic_words in topic_words_in]
topic_words_out = [sorted(topic_words,key=lambda x: x[1],reverse=True) for topic_words in topic_words_out]
log_writer = LogWriter(log_file_desc='LDATestsRegularize{}'.format(regularization))
log_writer.write_2D_list('topic_words_in', topic_words_in)
log_writer.write_2D_list('topic_words_out', topic_words_out)"""
topic_words_in_max = get_extremes(weight_in, num_of_topics,
num_of_important_words, reverse_word_map,
True, 'topic_words_in_max', log_writer,
dataset_helper.get_dataset_name())
topic_words_in_min = get_extremes(weight_in, num_of_topics,
num_of_important_words, reverse_word_map,
False, 'topic_words_in_min', log_writer,
dataset_helper.get_dataset_name())
result.append(datasets_helper.get_dataset_name())
model.summary(print_fn=result.append)
results.append(result)
results_saver.add_log("Done. Finishing this dataset.")
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss {}'.format(
datasets_helper.get_dataset_name()))
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.savefig(
results_saver.get_plot_path(datasets_helper.get_dataset_name(),
"loss"))
plt.clf()
acc = history.history['acc']
val_acc = history.history['val_acc']
plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy {}'.format(
datasets_helper.get_dataset_name()))
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.savefig(
results_saver.get_plot_path(datasets_helper.get_dataset_name(), "acc"))
plt.clf()
model.add(
Bidirectional(LSTM(enhanced_num_of_topics, return_sequences=True)))
#model.add(Bidirectional(LSTM(enhanced_num_of_topics, return_sequences=True)))
model.add(Bidirectional(LSTM(enhanced_num_of_topics)))
#model.add(Bidirectional(LSTM(enhanced_num_of_topics, activation='relu')))#,dropout=0.1,recurrent_dropout=0.5,input_shape=(num_of_words,),return_sequences=True))
#model.add(LSTM(40,activation='relu'))
#model.add(Dense(enhanced_num_of_topics, activation='relu', input_shape=(num_of_words,)))
#model.add(Dense(enhanced_num_of_topics, activation='relu'))
model.add(Dense(datasets_helper.get_num_of_topics(), activation='softmax'))
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
print(model.summary())
plot_model(model,
results_saver.get_plot_path("", "model-graph"),
show_shapes=True)
results_saver.add_log("Done. Now lets get training.")
batch_size = 512
history = model.fit(
x=Training_Text_Generator_RNN(datasets_helper.get_train_file_path(),
batch_size,
datasets_helper.get_num_of_train_texts(),
num_of_words, tokenizer, ";",
datasets_helper.get_num_of_topics()),
epochs=5,
validation_data=Training_Text_Generator_RNN(
datasets_helper.get_train_file_path(),
batch_size,
validation_count,
num_of_words,