In [2]:

    

# %load /home/sjkim/.jupyter/head.py
# %%writefile /home/sjkim/.jupyter/head.py
# %load /home/sjkim/.jupyter/head.py
%matplotlib inline
%load_ext autoreload 
%autoreload 2
from importlib import reload

import matplotlib.pyplot as plt
import numpy as np

import pandas as pd
import os


    

In [25]:

    

"""
LSTM example for Keras
======================
Object oriented style revision of the original code in Keras example
- https://github.com/fchollet/keras/blob/master/examples/imdb_lstm.py   
"""
from keras.preprocessing import sequence
from keras.datasets import imdb
from keras import layers, models


class Data:
    def __init__(self, max_features=20000, maxlen=80):
        (x_train, y_train), (x_test, y_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)
        
        self.x_train, self.y_train = x_train, y_train
        self.x_test, self.y_test = x_test, y_test
       
    
class RNN_LSTM(models.Model):
    def __init__(self, max_features, maxlen):
        x = layers.Input((maxlen,))
        h = layers.Embedding(max_features, 128)(x)
        h = layers.LSTM(128, dropout=0.2, recurrent_dropout=0.2)(h)
        y = layers.Dense(1, activation='sigmoid')(h)
        super().__init__(x, y)

        # try using different optimizers and different optimizer configs
        self.compile(loss='binary_crossentropy',
                      optimizer='adam', metrics=['accuracy'])        

class Machine:
    def __init__(self, 
                 max_features = 20000, 
                 maxlen = 80):
        self.data = Data(max_features, maxlen)
        self.model = RNN_LSTM(max_features, maxlen)
        
    def run(self, epochs=3, batch_size=32):
        data = self.data
        model = self.model
        print('Training stage')
        print('==============')
        model.fit(data.x_train, data.y_train,
                  batch_size=batch_size,
                  epochs=epochs,
                  validation_data=(data.x_test, data.y_test))
        score, acc = model.evaluate(data.x_test, data.y_test,
                                    batch_size=batch_size)
        print('Test performance: accuracy={0}, loss={1}'.format(acc, score))


    

In [26]:

    

m = Machine()
m.run()


    

    




Training stage
==============
Train on 25000 samples, validate on 25000 samples
Epoch 1/3
25000/25000 [==============================] - 126s - loss: 0.4605 - acc: 0.7826 - val_loss: 0.3689 - val_acc: 0.8386
Epoch 2/3
25000/25000 [==============================] - 126s - loss: 0.2974 - acc: 0.8790 - val_loss: 0.3930 - val_acc: 0.8358
Epoch 3/3
25000/25000 [==============================] - 125s - loss: 0.2144 - acc: 0.9188 - val_loss: 0.4274 - val_acc: 0.8130
25000/25000 [==============================] - 25s    
Test performance: accuracy=0.813, loss=0.42737488670825957

In [ ]: