IMDB with simple lstm and try with sequence autoencoder

try.py → imdb_lstm.py

 from keras.models import Sequential
 from keras.models import Sequential
-from keras.layers import Dense, Embedding, LSTM
+from keras.layers import Dense, Embedding, LSTM, Dropout
 from keras.datasets import imdb
 from keras.datasets import imdb
 from keras.preprocessing import sequence
 from keras.preprocessing import sequence
 import numpy as np
 import numpy as np
@@ -37,4 +37,4 @@ model.fit(x_train, y_train, batch_size=batch_size, epochs=15, validation_data=(x
 model.save('my_model.h5')
 model.save('my_model.h5')
 
 
 # print model's score and accuracy
 # print model's score and accuracy
-print('Score: %f\n Accuracy: %f' % model.evaluate(x_test, y_test, batch_size=batch_size))
+print('Score: %f\n Accuracy: %f' % model.evaluate(x_test, y_test, batch_size=batch_size))
\ No newline at end of file

imdb_lstm_autoencoder.py

+from numpy import array
+from keras.models import Sequential
+from keras.layers import TimeDistributed, Embedding, RepeatVector, Dense, LSTM
+from keras.utils import plot_model
+from keras.datasets import imdb
+from keras.preprocessing import sequence
+
+max_features = 10000
+maxlen = 9
+batch_size = 128
+
+(x_train, _), (x_test, _) = imdb.load_data(num_words=max_features)
+x_train = sequence.pad_sequences(x_train, maxlen=maxlen)
+x_test = sequence.pad_sequences(x_test, maxlen=maxlen)
+
+# reshape input into [samples, timesteps, features]
+timesteps = x_train.shape[1]
+samples = x_train.shape[0]
+x_train = x_train.reshape((samples, timesteps, 1))
+
+model = Sequential()
+# model.add(Embedding(max_features, 50, batch_input_shape=(batch_size, timesteps,1)))
+model.add(LSTM(512, activation='relu'))
+model.add(RepeatVector(timesteps))
+model.add(LSTM(512, activation='relu', return_sequences=True))
+model.add(TimeDistributed(Dense(1)))
+model.compile(optimizer='adam', loss='mse', metrics=['acc'])
+
+model.fit(x_train, x_train, epochs=300)
+print(model.evaluate(x_test, x_test))

imdb_simple_autoencoder.py

+import seq2seq
+from seq2seq.models import SimpleSeq2Seq
+from keras.datasets import imdb
+from keras.preprocessing import sequence
+import numpy as np
+from keras.layers import Input, Dense
+from keras.models import Model
+
+max_features = 10000
+maxlen = 400
+
+(x_train, _), (x_test, _) = imdb.load_data(num_words=max_features)
+x_train = sequence.pad_sequences(x_train, maxlen=maxlen)
+x_test = sequence.pad_sequences(x_test, maxlen=maxlen)
+
+x_train = np.reshape(x_train, (x_train.shape[0], 1, x_train.shape[1]))
+
+encoding_dim = 80
+
+input_seq = Input(shape=(None,maxlen))
+encoded = Dense(encoding_dim, activation='relu')(input_seq)
+decoded = Dense(maxlen, activation='sigmoid')(encoded)
+
+# this model maps an input to its reconstruction
+autoencoder = Model(input_seq, decoded)
+autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')
+
+autoencoder.fit(x_train, x_train, epochs=500, batch_size=128)
\ No newline at end of file