def test_unlabelled_classify():
if SUPERVISED == True:
outfile = './results/gsn_sup_test_outputs.csv'
with open("./results/gsn_sup_trained.pkl") as f:
gsn = pickle.load(f)
else:
outfile = './results/gsn_ae_test_outputs.csv'
with open("./results/gsn_ae_trained.pkl") as f:
gsn = pickle.load(f)
gsn = JointGSN.convert(gsn)
gsn._corrupt_switch = False
ds = MNIST(which_set='test',one_hot=True,all_labelled=ALL_LABELLED,supervised=SUPERVISED)
mean = gsn._get_aggregate_classification(ds.X)
am = np.argmax(mean, axis=1).astype(int)
print 'am shape: ', am.shape
test_output_file = open(outfile, "wb")
writer = csv.writer(test_output_file, delimiter=',')
writer.writerow(['Id', 'Prediction'])
for idx, predict in enumerate(am):
row = [idx+1, predict]
writer.writerow(row)
test_output_file.close()
def test_classify():
"""
See how well a (supervised) GSN performs at classification.
"""
with open("gsn_sup_example.pkl") as f:
gsn = pickle.load(f)
gsn = JointGSN.convert(gsn)
# turn off corruption
gsn._corrupt_switch = False
ds = MNIST(which_set='test', one_hot=True)
mb_data = ds.X
y = ds.y
for i in xrange(1, 10):
y_hat = gsn.classify(mb_data, trials=i)
errors = np.abs(y_hat - y).sum() / 2.0
# error indices
#np.sum(np.abs(y_hat - y), axis=1) != 0
print i, errors, errors / mb_data.shape[0]
def test_classify():
with open("./results/gsn_sup_trained.pkl") as f:
gsn = pickle.load(f)
gsn = JointGSN.convert(gsn)
gsn._corrupt_switch = False
ds = MNIST(which_set='test',one_hot=True,all_labelled=ALL_LABELLED,supervised=SUPERVISED)
mb_data = ds.X
y = ds.y
outfile = open("./results/gsn_train_outputs.csv","wb")
writer = csv.writer(outfile, delimiter=',', quotechar='"', quoting=csv.QUOTE_ALL)
for i in xrange(1, 10):
y_hat = gsn.classify(mb_data, trials=i)
errors = np.abs(y_hat - y).sum() / 2.0
errors_normalize = errors / mb_data.shape[0]
writer.writerow([i, errors, errors_normalize])
writer.writerow(y_hat)
print i, errors, errors_normalize
outfile.close()
def test_classify():
"""
See how well a (supervised) GSN performs at classification.
"""
with open("gsn_sup_example.pkl") as f:
gsn = pickle.load(f)
gsn = JointGSN.convert(gsn)
# turn off corruption
gsn._corrupt_switch = False
ds = MNIST(which_set='test')
mb_data = ds.X
y = ds.y
for i in xrange(1, 10):
y_hat = gsn.classify(mb_data, trials=i)
errors = np.abs(y_hat - y).sum() / 2.0
# error indices
#np.sum(np.abs(y_hat - y), axis=1) != 0
print(i, errors, errors / mb_data.shape[0])
