def log(self, msg, color=None):
if color is None:
if self.verbose:
logger.info(msg)
else:
logger.info(colored("{}".format(msg), color))
output_size=N_HIDDENS,
learning_rate=0.01,
max_depth=3,
num_boost_round=50)
G = MultiXGBModel(input_size=N_HIDDENS,
output_size=3,
learning_rate=0.01,
max_depth=3,
num_boost_round=50)
Ginv = MultiXGBModel(input_size=3,
output_size=N_HIDDENS,
learning_rate=0.01,
max_depth=3,
num_boost_round=50)
net = MGBDTAntoEncoder(F, G, Ginv, target_lr=0.05, epsilon=0.3)
logger.info("net architecture")
logger.info(net)
net.init(X, n_rounds=50)
# net.init(X, n_rounds=5)
net.fit(X, X, n_epochs=100)
hiddens = net.get_hiddens(X)
for i in range(0, N_HIDDENS):
for j in range(i + 1, N_HIDDENS):
des_path = osp.join(des_dir,
"pred1.{}_{}.jpg".format(i + 1, j + 1))
plot2d(hiddens[1][:, [i, j]], color=color, save_path=des_path)
save_path = osp.join(des_dir, "pred2.jpg")
plot3d(hiddens[2], color=color, save_path=save_path)
learning_rate=0.1,
max_depth=5,
num_boost_round=5))
net.add_layer("tp_layer",
F=MultiXGBModel(input_size=3,
output_size=2,
learning_rate=0.1,
max_depth=5,
num_boost_round=5),
G=MultiXGBModel(input_size=2,
output_size=3,
learning_rate=0.1,
max_depth=5,
num_boost_round=5))
logger.info(net)
net.init(x_train, n_rounds=5)
net.fit(x_train,
y_train,
n_epochs=50,
eval_sets=[(x_test, y_test)],
eval_metric="accuracy")
colors = np.asarray(["red", "blue"])
# save the input dataset
save_path = osp.join(des_dir, "input.jpg")
plot2d(x_test, color=colors[y_test], save_path=save_path)
# save the hidden output
save_path = osp.join(des_dir, "pred2.jpg")
hiddens = net.get_hiddens(x_test)
import sys
sys.path.insert(0, "lib")
from mgbdt import MGBDT
from mgbdt import MultiXGBModel, LinearModel
from mgbdt.datasets import uci_adult
from mgbdt.utils.exp_utils import set_seed
from mgbdt.utils.log_utils import logger
if __name__ == "__main__":
set_seed(666)
(x_train, y_train), (x_test, y_test) = uci_adult.load_data()
logger.info("x_train.shape={}, y_train.shape={}".format(
x_train.shape, y_train.shape))
logger.info("x_test.shape={}, y_test.shape={}".format(
x_test.shape, y_test.shape))
net = MGBDT(loss=None, target_lr=1.0, epsilon=0.3)
net.add_layer("tp_layer",
F=MultiXGBModel(input_size=113,
output_size=128,
learning_rate=0.3,
max_depth=5,
num_boost_round=5),
G=None)
net.add_layer("tp_layer",
F=MultiXGBModel(input_size=128,
output_size=128,
learning_rate=0.3,
max_depth=5,
parser.add_argument("--n_layers", dest="n_layers", default=3, type=int)
args = parser.parse_args()
return args
if __name__ == "__main__":
args = parse_args()
n_layers = args.n_layers
set_seed(0)
X, y = uci_yeast.load_data()
n_splits = 10
skf = StratifiedKFold(n_splits=n_splits, random_state=0, shuffle=True)
accs = []
for ci, (train_index, test_index) in enumerate(skf.split(X, y)):
logger.info("[progress] cv={}/{}".format(ci + 1, n_splits))
x_train, x_test = X[train_index], X[test_index]
y_train, y_test = y[train_index], y[test_index]
logger.info("x_train.shape={}, y_train.shape={}".format(
x_train.shape, y_train.shape))
logger.info("x_test.shape={}, y_test.shape={}".format(
x_test.shape, y_test.shape))
net = MGBDT(loss=None, target_lr=1.0, epsilon=0.3)
net.add_layer("tp_layer",
F=MultiXGBModel(input_size=8,
output_size=16,
learning_rate=0.1,
max_depth=5,
num_boost_round=5),
G=None)