LSTM Binary classification with Keras

input.csv

LSTM_Binary.py

from keras.layers import Dense, Dropout, LSTM, Embedding
from keras.preprocessing.sequence import pad_sequences
from keras.models import Sequential
import pandas as pd
import numpy as np

input_file = 'input.csv'

def load_data(test_split = 0.2):
    print ('Loading data...')
    df = pd.read_csv(input_file)
    df['sequence'] = df['sequence'].apply(lambda x: [int(e) for e in x.split()])
    df = df.reindex(np.random.permutation(df.index))

    train_size = int(len(df) * (1 - test_split))

    X_train = df['sequence'].values[:train_size]
    y_train = np.array(df['target'].values[:train_size])
    X_test = np.array(df['sequence'].values[train_size:])
    y_test = np.array(df['target'].values[train_size:])

    return pad_sequences(X_train), y_train, pad_sequences(X_test), y_test


def create_model(input_length):
    print ('Creating model...')
    model = Sequential()
    model.add(Embedding(input_dim = 188, output_dim = 50, input_length = input_length))
    model.add(LSTM(output_dim=256, activation='sigmoid', inner_activation='hard_sigmoid', return_sequences=True))
    model.add(Dropout(0.5))
    model.add(LSTM(output_dim=256, activation='sigmoid', inner_activation='hard_sigmoid'))
    model.add(Dropout(0.5))
    model.add(Dense(1, activation='sigmoid'))

    print ('Compiling...')
    model.compile(loss='binary_crossentropy',
                  optimizer='rmsprop',
                  metrics=['accuracy'])
    return model


X_train, y_train, X_test, y_test = load_data()

model = create_model(len(X_train[0]))

print ('Fitting model...')
hist = model.fit(X_train, y_train, batch_size=64, nb_epoch=10, validation_split = 0.1, verbose = 1)

score, acc = model.evaluate(X_test, y_test, batch_size=1)
print('Test score:', score)
print('Test accuracy:', acc)

hello,i use your code,and implement it,but it has errors:InternalError (see above for traceback): Blas GEMM launch failed : a.shape=(24, 50), b.shape=(50, 256), m=24, n=256, k=50

so do you have solutions?

If the output was string value, Is it possible that classify our data?

I tried:
y_pred = model.predict(X_test)

But Im not getting the target

would it work if inputs are string values, like date - '03/07/2012' ?Thanks.

Hey, this example does not learn, it only returns 0, no matter what sequence.

@guysoft, Did you find the solution to the problem? I am also having the same issue. I tried to print out the gradients to see if there was any gradient flow as described : https://gist.github.com/mickypaganini/a2291691924981212b4cfc8e600e52b1 , but was having issue with that as well.

what to do if the sequences have negative values as well?

what to do if the sequences have negative values as well?

If you are still looking for a solution,
1)Replace every negative sign with a 0. Eg- 2-31=2031 or 12-6=1206. This will work correctly if your sequence itself does not involve zeros.
2) or alternatively, convert the sequence into a binary representation.