print("Data set loaded...")
# Sample size
n_samples = len(data_set['words'])
fold_size = int(math.ceil(n_samples / 10.0))
# Get truths
truths = []
for truth in data_set['labels']:
truths += [truth[0]]
# end for
# Deep-Learning model
deep_learning_model = PAN17DeepNNModel(PAN17ConvNet(n_classes=2),
classes=("male", "female"),
cuda=args.cuda,
lr=args.lr,
momentum=args.momentum,
log_interval=args.log_interval,
seed=args.seed)
# K-10 fold
grams_set = np.array(data_set['words'])
truths_set = np.array(truths)
grams_set.shape = (10, fold_size)
truths_set.shape = (10, fold_size)
# Select training and test sets
test = grams_set[-1]
test_truths = truths_set[-1]
training = np.delete(grams_set, -1, axis=0)
training_truths = np.delete(truths_set, -1, axis=0)
data_set = pickle.load(f)
# Sample size
n_samples = len(data_set['2grams'])
fold_size = int(math.ceil(n_samples / 10.0))
# Get truths
truths = []
for truth in data_set['labels']:
truths += [truth[1]]
# end for
# Deep-Learning model
deep_learning_model = PAN17DeepNNModel(PAN17ConvNet(n_classes=5), classes=('great britain', 'canada',
'ireland', 'new zealand',
'australia'), cuda=args.cuda,
lr=args.lr, momentum=args.momentum, log_interval=args.log_interval,
seed=args.seed)
# K-10 fold
grams_set = np.array(data_set['2grams'])
m_height = grams_set.shape[1]
m_width = grams_set.shape[2]
truths_set = np.array(truths)
grams_set.shape = (10, fold_size, m_height, m_width)
truths_set.shape = (10, fold_size)
# Select training and test sets
test = grams_set[-1]
test_truths = truths_set[-1]
training = np.delete(grams_set, -1, axis=0)
# Get truths
truths = []
for truth in data_set['labels']:
truths += [truth[1]]
# end for
minis = np.zeros(args.k)
success_rates = np.zeros((args.epoch, args.k))
for k in range(args.k):
print("%d th fold" % k)
# Deep-Learning model
deep_learning_model = PAN17DeepNNModel(PAN17ConvNet(n_classes=len(params[args.lang][2]), params=(params[args.lang][0],
params[args.lang][1])), classes=params[args.lang][2], cuda=args.cuda, lr=args.lr,
momentum=args.momentum, log_interval=args.log_interval,
seed=args.seed)
# K-10 fold
grams_set = np.array(data_set['2grams'])
m_height = grams_set.shape[1]
m_width = grams_set.shape[2]
truths_set = np.array(truths)
grams_set.shape = (10, fold_size, m_height, m_width)
truths_set.shape = (10, fold_size)
# Select training and test sets
test = grams_set[k]
test_truths = truths_set[k]
training = np.delete(grams_set, k, axis=0)
training_truths = np.delete(truths_set, k, axis=0)