lstm.py

import numpy as np
from sklearn.cross_validation import train_test_split
from sklearn import preprocessing


data = np.load("data/data.npz")

X_train, X_test, y_train, y_test, w_train, w_test = train_test_split(data['X'], data['y'], data['w'], test_size=0.33, random_state=42)

import logging
from sklearn.metrics import roc_auc_score
from keras.callbacks import Callback

class IntervalEvaluation(Callback):
    def __init__(self, validation_data=(), interval=1):
        super(Callback, self).__init__()

        self.interval = interval
        self.X_val, self.y_val, self.w_val = validation_data

    def on_epoch_end(self, epoch, logs={}):
        if epoch % self.interval == 0:
            y_pred = self.model.predict_proba(self.X_val, batch_size=16000,verbose=0)
            score = roc_auc_score(self.y_val, y_pred, sample_weight=self.w_val)
            print "interval evaluation - epoch: {:d} - score: {:.6f} \n".format(epoch, score)
            
ival = IntervalEvaluation(validation_data=(X_test, y_test,w_test), interval=1)

from keras.models import Sequential
from keras.layers.core import Masking
from keras.layers import LSTM,Dense, Dropout, Activation

model = Sequential()

model.add(Masking(mask_value=0.,input_shape=X_train[0].shape))
model.add(LSTM(256)) 
model.add(Dense(1024, activation="relu"))
model.add(Dense(1024, activation="relu"))
model.add(Dense(1))
model.add(Activation('sigmoid'))

model.compile(loss='binary_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])

model.fit(X_train, y_train, batch_size=256, nb_epoch=10, validation_data = (X_test, y_test),callbacks=[ival], class_weight=[13.585,1])