In [12]:

    

import numpy as np
import pandas as pd
import timeit

from keras.models import Sequential
from keras.layers.core import Dense

train = pd.read_csv('/home/jake/kaggle/liberty_mutual_property_inspection/input/train_dummies.csv')
test = pd.read_csv('/home/jake/kaggle/liberty_mutual_property_inspection/input/test_dummies.csv')

X_train = train.drop('Hazard', axis=1).values.astype(np.float32)/255
y_train = train.Hazard.values.astype(np.float32).reshape(-1, 1)
X_test = test.values.astype(np.float32)/255

in_dim = X_train.shape[1]
out_dim = 1


    

In [25]:

    

t0 = timeit.default_timer()

model = Sequential()
model.add(Dense(in_dim, 256, activation='relu'))
model.add(Dense(256, 256, activation='relu'))
model.add(Dense(256, out_dim))
model.compile(loss='mse', optimizer='rmsprop')

print('Build time: {:.3f} s'.format(timeit.default_timer()-t0))


    

    




Build time: 2.036 s

In [26]:

    

model.fit(X_train, y_train, nb_epoch=5, batch_size=128)


    

    




Epoch 0
50999/50999 [==============================] - 12s - loss: 52.6942    
Epoch 1
50999/50999 [==============================] - 11s - loss: 19.0833    
Epoch 2
50999/50999 [==============================] - 11s - loss: 17.0826    
Epoch 3
50999/50999 [==============================] - 11s - loss: 16.6745    
Epoch 4
50999/50999 [==============================] - 11s - loss: 16.5212    






    
Out[26]:





<keras.callbacks.History at 0x7f9e5cde0320>

In [ ]: