def do_fold_libsvm(bs = 2**15):
# Low memory
print('Loading')
read = pd.HDFStore('data/train_folds.h5', mode='r')
for i in xrange(10):
print('Fold {}'.format(i))
print('Training set')
name = 'train_fold_{}'.format(i)
nrows = read.get_storer(name).nrows
pb = progressbar.ProgressBar(maxval=nrows//bs).start()
with open('train_folds_{}.txt'.format(i), 'wb') as out:
for j,chunk in enumerate(read.select(name, iterator=True, chunksize=bs)):
pb.update(j)
X = chunk.drop('Expected', axis=1).values.astype('float')
y = chunk['Expected'].values.clip(0,70).astype('int')
RainCompetition.to_libsvm(X, y, out)
pb.finish()
print('Validation set')
name = 'valid_fold_{}'.format(i)
nrows = read.get_storer(name).nrows
pb = progressbar.ProgressBar(maxval=nrows//bs).start()
with open('valid_folds_{}.txt'.format(i), 'wb') as out:
for j,chunk in enumerate(read.select(name, iterator=True, chunksize=bs)):
pb.update(j)
X = chunk.drop('Expected', axis=1).values.astype('float')
y = chunk['Expected'].values.clip(0,70).astype('int')
RainCompetition.to_libsvm(X, y, out)
pb.finish()
('dense1', DenseLayer),
('dropout1', DropoutLayer),
('dense2', DenseLayer),
('dropout2', DropoutLayer),
('dense3', DenseLayer),
('output', DenseLayer)],
input_shape=(None, num_features),
dense1_num_units=512,
dropout1_p=0.5,
dense2_num_units=512,
dropout2_p=0.5,
dense3_num_units=512,
output_num_units=num_classes,
output_nonlinearity=softmax,
update=nesterov_momentum,
eval_size=0.2,
verbose=1,
update_learning_rate=theano.shared(float32(0.01)),
update_momentum=theano.shared(float32(0.9)),
on_epoch_finished=[
AdjustVariable('update_learning_rate', start=0.01, stop=0.00001),
AdjustVariable('update_momentum', start=0.9, stop=0.999),
EarlyStopping(),
],
max_epochs=10000,)
net0.initialize()
# do_fit(net0, 'data/train_impu_norm_shuf.csv', n_iter=1)
net0.load_weights_from('nn_weights')
RainCompetition.do_predict(net0, RainCompetition.__data__['test_normalized'], 'data/rain_nn_pred.csv')