def year():
executor = HyperParameterSearch(pipeline_cls=HoldoutPipeline,
objective_cls=MseObjective,
output_dir=ex.artifacts_dir)
config = dict(dataset=FileDataset(file_path='data/yearmsd.csv',
standardize=True,
shuffle=False),
split=dict(train_size=0.9, ),
method=PiEnsemble,
hyper_params=dict(
ensemble_size=5,
aggreg_func=no_aggreg,
hidden_size=[100, 100],
epochs=5000,
batch_size=1000,
optimizer=Adam,
learning_rate=lambda t: t.suggest_discrete_uniform(
'learning_rate', 0.001, 0.01, 0.001),
scheduler=ExponentialDecay,
decay_rate=lambda t: t.suggest_discrete_uniform(
'decay_rate', 0.95, 1., 0.01),
decay_steps=50.,
early_stopping=True,
punish_crossing=False,
patience=100,
delta=1e-6,
tolerance=0.01,
loss_func=mse_loss,
alpha=None,
print_frequency=10,
metrics=[mse],
device='cpu'),
num_trials=300)
def year():
config = dict(
dataset=FileDataset(
file_path='data/yearmsd.csv',
standardize=True,
shuffle=False
),
split=dict(
train_size=0.9,
test_size=0.1
),
num_runs=1,
method=PiEnsemble,
hyper_params=dict( # HPS trial number 47
ensemble_size=5,
aggreg_func=mean_aggreg,
hidden_size=[100, 100],
epochs=4,
batch_size=1000,
optimizer=Adam,
learning_rate=0.009,
scheduler=ExponentialDecay,
decay_rate=0.95,
decay_steps=50.,
loss_func=mse_loss,
alpha=None,
metrics=[mse],
print_frequency=10,
device='cpu'
)
)
def year():
config = dict(
dataset=FileDataset(file_path='data/yearmsd.csv',
standardize=True,
shuffle=False),
split=dict(train_size=0.9, test_size=0.1),
num_runs=1,
method=PiEnsemble,
hyper_params=dict( # from HPS trial 77 and then manually fine-tuned
ensemble_size=5,
aggreg_func=[sem_aggreg, std_aggreg, snm_aggreg],
hidden_size=[100, 100],
optimizer=Adam,
learning_rate=0.005,
scheduler=ExponentialDecay,
decay_steps=50.,
decay_rate=0.99,
epochs=40,
batch_size=1000,
loss_func=qd_plus_loss,
alpha=0.05,
soften=160.,
lambda_1=0.999,
lambda_2=0.1,
ksi=10.,
print_frequency=1,
device='cpu'))
def year():
config = dict(
dataset=FileDataset(
file_path='data/yearmsd.csv',
standardize=True,
shuffle=False
),
split=dict(
train_size=0.9,
test_size=0.1
),
num_runs=1,
method=MvEnsemble,
hyper_params=dict( # HPS trial number 2
ensemble_size=5,
aggreg_func=[mv_aggreg],
hidden_size=[100, 100],
epochs=4,
batch_size=100,
optimizer=Adam,
learning_rate=0.004,
scheduler=ExponentialDecay,
decay_rate=.99,
decay_steps=50.,
loss_func=normal_loss,
epsilon=None, # `None` to disable adversarial examples
alpha=0.05,
print_frequency=1,
device='cpu'
)
)
def year():
config = dict(
dataset=FileDataset(file_path='data/yearmsd.csv',
standardize=True,
shuffle=False),
split=dict(
train_size=0.9,
val_size=0.,
test_size=0.1,
),
num_runs=1,
method=PiEnsemble,
hyper_params=dict( # HPS trial number 206
ensemble_size=5,
aggreg_func=[sem_aggreg, std_aggreg],
hidden_size=[100, 100],
optimizer=Adam,
learning_rate=0.001,
scheduler=ExponentialDecay,
decay_steps=50.,
decay_rate=1.0,
epochs=64,
batch_size=1000,
loss_func=qd_code_loss,
alpha=0.05,
soften=160.,
lambda_=11.,
print_frequency=10,
device='cpu'))
def year_paper():
config = dict(dataset=FileDataset(file_path='data/yearmsd.csv',
standardize=True,
shuffle=False),
split=dict(train_size=0.9, test_size=0.1),
num_runs=1,
method=PiEnsemble,
hyper_params=dict(ensemble_size=5,
aggreg_func=[sem_aggreg, std_aggreg],
hidden_size=[100],
optimizer=Adam,
learning_rate=0.005,
scheduler=ExponentialDecay,
decay_steps=50.,
decay_rate=0.999,
epochs=100,
batch_size=1000,
loss_func=qd_paper_loss,
retry_on_crossing=False,
alpha=0.05,
soften=160.,
lambda_=15.,
print_frequency=10))