Credits: Forked from deep-learning-keras-tensorflow by Valerio Maggio

Recurrent Neural networks

RNN

<img src ="imgs/rnn.png" width="20%">

A recurrent neural network (RNN) is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior.

In [ ]:
keras.layers.recurrent.SimpleRNN(output_dim, 
                                 init='glorot_uniform', inner_init='orthogonal', activation='tanh', 
                                 W_regularizer=None, U_regularizer=None, b_regularizer=None, 
                                 dropout_W=0.0, dropout_U=0.0)

Backprop Through time

Contrary to feed-forward neural networks, the RNN is characterized by the ability of encoding longer past information, thus very suitable for sequential models. The BPTT extends the ordinary BP algorithm to suit the recurrent neural architecture.

<img src ="imgs/rnn2.png" width="45%">

In [3]:
%matplotlib inline
In [1]:
import numpy as np
import pandas as pd
import theano
import theano.tensor as T
import keras 
from keras.models import Sequential
from keras.layers import Dense, Activation
from keras.preprocessing import image
from __future__ import print_function
import numpy as np
import matplotlib.pyplot as plt

from keras.datasets import imdb
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D
from keras.utils import np_utils
from keras.preprocessing import sequence
from keras.layers.embeddings import Embedding
from keras.layers.recurrent import LSTM, GRU, SimpleRNN
from sklearn.preprocessing import LabelEncoder
from sklearn.preprocessing import StandardScaler
from sklearn.cross_validation import train_test_split
from keras.layers.core import Activation, TimeDistributedDense, RepeatVector
from keras.callbacks import EarlyStopping, ModelCheckpoint
Using Theano backend.

IMDB sentiment classification task

This is a dataset for binary sentiment classification containing substantially more data than previous benchmark datasets.

IMDB provided a set of 25,000 highly polar movie reviews for training, and 25,000 for testing.

There is additional unlabeled data for use as well. Raw text and already processed bag of words formats are provided.

http://ai.stanford.edu/~amaas/data/sentiment/

Data Preparation - IMDB

In [4]:
max_features = 20000
maxlen = 100  # cut texts after this number of words (among top max_features most common words)
batch_size = 32

print("Loading data...")
(X_train, y_train), (X_test, y_test) = imdb.load_data(nb_words=max_features, test_split=0.2)
print(len(X_train), 'train sequences')
print(len(X_test), 'test sequences')

print('Example:')
print(X_train[:1])

print("Pad sequences (samples x time)")
X_train = sequence.pad_sequences(X_train, maxlen=maxlen)
X_test = sequence.pad_sequences(X_test, maxlen=maxlen)
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
Loading data...
20000 train sequences
5000 test sequences
Example:
[ [1, 20, 28, 716, 48, 495, 79, 27, 493, 8, 5067, 7, 50, 5, 4682, 13075, 10, 5, 852, 157, 11, 5, 1716, 3351, 10, 5, 500, 7308, 6, 33, 256, 41, 13610, 7, 17, 23, 48, 1537, 3504, 26, 269, 929, 18, 2, 7, 2, 4284, 8, 105, 5, 2, 182, 314, 38, 98, 103, 7, 36, 2184, 246, 360, 7, 19, 396, 17, 26, 269, 929, 18, 1769, 493, 6, 116, 7, 105, 5, 575, 182, 27, 5, 1002, 1085, 130, 62, 17, 24, 89, 17, 13, 381, 1421, 8, 5167, 7, 5, 2723, 38, 325, 7, 17, 23, 93, 9, 156, 252, 19, 235, 20, 28, 5, 104, 76, 7, 17, 169, 35, 14764, 17, 23, 1460, 7, 36, 2184, 934, 56, 2134, 6, 17, 891, 214, 11, 5, 1552, 6, 92, 6, 33, 256, 82, 7]]
Pad sequences (samples x time)
X_train shape: (20000L, 100L)
X_test shape: (5000L, 100L)

Model building

In [7]:
print('Build model...')
model = Sequential()
model.add(Embedding(max_features, 128, input_length=maxlen))
model.add(SimpleRNN(128))  
model.add(Dropout(0.5))
model.add(Dense(1))
model.add(Activation('sigmoid'))

# try using different optimizers and different optimizer configs
model.compile(loss='binary_crossentropy', optimizer='adam', class_mode="binary")

print("Train...")
model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=1, validation_data=(X_test, y_test), show_accuracy=True)
Build model...
Train...
Train on 20000 samples, validate on 5000 samples
Epoch 1/1
20000/20000 [==============================] - 174s - loss: 0.7213 - val_loss: 0.6179
Out[7]:
<keras.callbacks.History at 0x20519860>

LSTM

A LSTM network is an artificial neural network that contains LSTM blocks instead of, or in addition to, regular network units. A LSTM block may be described as a "smart" network unit that can remember a value for an arbitrary length of time.

Unlike traditional RNNs, an Long short-term memory network is well-suited to learn from experience to classify, process and predict time series when there are very long time lags of unknown size between important events.

<img src ="imgs/gru.png" width="60%">

In [ ]:
keras.layers.recurrent.LSTM(output_dim, init='glorot_uniform', inner_init='orthogonal', 
                            forget_bias_init='one', activation='tanh', 
                            inner_activation='hard_sigmoid', 
                            W_regularizer=None, U_regularizer=None, b_regularizer=None, 
                            dropout_W=0.0, dropout_U=0.0)

GRU

Gated recurrent units are a gating mechanism in recurrent neural networks.

Much similar to the LSTMs, they have fewer parameters than LSTM, as they lack an output gate.

In [ ]:
keras.layers.recurrent.GRU(output_dim, init='glorot_uniform', inner_init='orthogonal', 
                           activation='tanh', inner_activation='hard_sigmoid', 
                           W_regularizer=None, U_regularizer=None, b_regularizer=None, 
                           dropout_W=0.0, dropout_U=0.0)

Your Turn! - Hands on Rnn

In [ ]:
print('Build model...')
model = Sequential()
model.add(Embedding(max_features, 128, input_length=maxlen))

# Play with those! try and get better results!
#model.add(SimpleRNN(128))  
#model.add(GRU(128))  
#model.add(LSTM(128))  

model.add(Dropout(0.5))
model.add(Dense(1))
model.add(Activation('sigmoid'))

# try using different optimizers and different optimizer configs
model.compile(loss='binary_crossentropy', optimizer='adam', class_mode="binary")

print("Train...")
model.fit(X_train, y_train, batch_size=batch_size, 
          nb_epoch=4, validation_data=(X_test, y_test), show_accuracy=True)
score, acc = model.evaluate(X_test, y_test, batch_size=batch_size, show_accuracy=True)
print('Test score:', score)
print('Test accuracy:', acc)

Sentence Generation using RNN(LSTM)

In [ ]:
from keras.models import Sequential
from keras.layers import Dense, Activation, Dropout
from keras.layers import LSTM
from keras.optimizers import RMSprop
from keras.utils.data_utils import get_file
import numpy as np
import random
import sys

path = get_file('nietzsche.txt', origin="https://s3.amazonaws.com/text-datasets/nietzsche.txt")
text = open(path).read().lower()
print('corpus length:', len(text))

chars = sorted(list(set(text)))
print('total chars:', len(chars))
char_indices = dict((c, i) for i, c in enumerate(chars))
indices_char = dict((i, c) for i, c in enumerate(chars))

# cut the text in semi-redundant sequences of maxlen characters
maxlen = 40
step = 3
sentences = []
next_chars = []
for i in range(0, len(text) - maxlen, step):
    sentences.append(text[i: i + maxlen])
    next_chars.append(text[i + maxlen])
print('nb sequences:', len(sentences))

print('Vectorization...')
X = np.zeros((len(sentences), maxlen, len(chars)), dtype=np.bool)
y = np.zeros((len(sentences), len(chars)), dtype=np.bool)
for i, sentence in enumerate(sentences):
    for t, char in enumerate(sentence):
        X[i, t, char_indices[char]] = 1
    y[i, char_indices[next_chars[i]]] = 1


# build the model: a single LSTM
print('Build model...')
model = Sequential()
model.add(LSTM(128, input_shape=(maxlen, len(chars))))
model.add(Dense(len(chars)))
model.add(Activation('softmax'))

optimizer = RMSprop(lr=0.01)
model.compile(loss='categorical_crossentropy', optimizer=optimizer)


def sample(preds, temperature=1.0):
    # helper function to sample an index from a probability array
    preds = np.asarray(preds).astype('float64')
    preds = np.log(preds) / temperature
    exp_preds = np.exp(preds)
    preds = exp_preds / np.sum(exp_preds)
    probas = np.random.multinomial(1, preds, 1)
    return np.argmax(probas)

# train the model, output generated text after each iteration
for iteration in range(1, 60):
    print()
    print('-' * 50)
    print('Iteration', iteration)
    model.fit(X, y, batch_size=128, nb_epoch=1)

    start_index = random.randint(0, len(text) - maxlen - 1)

    for diversity in [0.2, 0.5, 1.0, 1.2]:
        print()
        print('----- diversity:', diversity)

        generated = ''
        sentence = text[start_index: start_index + maxlen]
        generated += sentence
        print('----- Generating with seed: "' + sentence + '"')
        sys.stdout.write(generated)

        for i in range(400):
            x = np.zeros((1, maxlen, len(chars)))
            for t, char in enumerate(sentence):
                x[0, t, char_indices[char]] = 1.

            preds = model.predict(x, verbose=0)[0]
            next_index = sample(preds, diversity)
            next_char = indices_char[next_index]

            generated += next_char
            sentence = sentence[1:] + next_char

            sys.stdout.write(next_char)
            sys.stdout.flush()
        print()
Downloading data from https://s3.amazonaws.com/text-datasets/nietzsche.txt
598016/600901 [============================>.] - ETA: 0s('corpus length:', 600901)
('total chars:', 59)
('nb sequences:', 200287)
Vectorization...
Build model...
()
--------------------------------------------------
('Iteration', 1)
Epoch 1/1
200287/200287 [==============================] - 1367s - loss: 1.9977  
()
('----- diversity:', 0.2)
----- Generating with seed: "nd the frenzied
speeches of the prophets"
nd the frenzied
speeches of the prophets and the present and and the preases and the soul to the sense of the morals and the some the consequence of the most and one only the some of the proment and interent of the some devertal to the self-consertion of the some deverent of the some distiness and the sense of the some of the morality of the most proves and the some of the some in the seem of the self-conception of the sees of the sense()
()
('----- diversity:', 0.5)
----- Generating with seed: "nd the frenzied
speeches of the prophets"
nd the frenzied
speeches of the prophets of the preat weak to the master of man who onow in interervain of even which who with it is the isitaial conception of the some live the contented the one who exilfacied in the sees to raters, and the passe expecience the inte that the persented in the pass, in the experious of the soulity of the waith the morally distanding of the some of the most interman only and as a period of the sense and o()
()
('----- diversity:', 1.0)
----- Generating with seed: "nd the frenzied
speeches of the prophets"
nd the frenzied
speeches of the prophets of
ar self now no ecerspoped ivent so not,
that itsed undiswerbatarlials. what it is altrenively evok
now be scotnew
prigardiness intagualds, and coumond-grow to
the respence you as penires never wand be
natuented ost ablinice to love worts an who itnopeancew be than mrank againribl
some something lines in the estlenbtupenies of korils divenowry apmains, curte, were,
ind "feulness.  a will, natur()
()
('----- diversity:', 1.2)
----- Generating with seed: "nd the frenzied
speeches of the prophets"
nd the frenzied
speeches of the prophets, ind someaterting will stroour hast-fards and lofe beausold, in souby in ruarest, we withquus. "the capinistin and it a mode what it be
my oc, to th[se condectay
of ymo fre
dunt and so asexthersess renieved concecunaulies tound"), from glubiakeitiouals kenty am feelitafouer deceanw or sumpind, and by afolod peall--phasoos of sole
iy copprajakias
in
in adcyont-mean to prives apf-rigionall thust wi()
()
--------------------------------------------------
('Iteration', 2)
Epoch 1/1
 40576/200287 [=====>........................] - ETA: 1064s - loss: 1.6878