def apply_transform(features_DC,
features_IC,
output_labels=None,
energy_factor=100.,
track_factor=200.,
transform="MaxAbs",
transform_output=True):
from scaler_transformations import TransformData, new_transform
static_stats = [25., 4000., 4000., 4000., 2000.]
low_stat_DC = static_stats
high_stat_DC = static_stats
low_stat_IC = static_stats
high_stat_IC = static_stats
input_transform_factors = np.array([high_stat_DC, high_stat_IC])
features_DC = new_transform(features_DC)
features_DC = TransformData(features_DC,
low_stats=low_stat_DC,
high_stats=high_stat_DC,
scaler=transform)
features_IC = new_transform(features_IC)
features_IC = TransformData(features_IC,
low_stats=low_stat_IC,
high_stats=high_stat_IC,
scaler=transform)
output_label_names = np.array([
"Energy", "Zenith", "Azimuth", "Time", "X", "Y", "Z", "Track",
"IsTrack", "Flavor", "IsAntineutrino", "IsCC"
])
output_names = np.array(output_label_names)
output_transform_factors = np.ones((len(output_names)))
if transform_output:
# switch track and azimuth positions
track = np.copy(output_labels[:, 7])
azimuth = np.copy(output_labels[:, 2])
#Make changes to output_labels array
output_labels[:,
0] = output_labels[:, 0] / float(energy_factor) #energy
output_labels[:, 1] = np.cos(output_labels[:, 1]) #cos zenith
output_labels[:,
2] = track / float(track_factor) #MAKE TRACK THIRD INPUT
output_labels[:, 7] = azimuth #MOVE AZIMUTH TO WHERE TRACK WAS
#Track changes in output arrarys
output_label_names[1] = "Cosine Zenith"
output_label_names[2] = "Track"
output_label_names[7] = "Azimuth"
output_transform_factors[0] = energy_factor
output_transform_factors[2] = track_factor
return features_DC, features_IC, output_labels, output_label_names, input_transform_factors, output_transform_factors
def apply_transform(features_DC, features_IC, labels=None, energy_factor=100., track_factor=200.,transform="MaxAbs"):
from scaler_transformations import TransformData, new_transform
static_stats = [25., 4000., 4000., 4000., 2000.]
low_stat_DC = static_stats
high_stat_DC = static_stats
low_stat_IC = static_stats
high_stat_IC = static_stats
features_DC = new_transform(features_DC)
features_DC = TransformData(features_DC, low_stats=low_stat_DC, high_stats=high_stat_DC, scaler=transform)
features_IC = new_transform(features_IC)
features_IC = TransformData(features_IC, low_stats=low_stat_IC, high_stats=high_stat_IC, scaler=transform)
return features_DC, features_IC
full_reco=full_reco, full_initial_stats=full_initial_stats,\
full_num_pulses=full_num_pulses,full_weights=full_weights,use_old_reco_flag=use_old_reco)
# Check that it follows the usual energy and CC cuts
over_emax = full_labels[:, 0] > emax
under_emin = full_labels[:, 0] < emin
assert sum(over_emax) == 0, "Have events greater than emax in final sample"
assert sum(under_emin) == 0, "Have events less than emin in final sample"
if cut_name == "CC":
isCC = full_labels[:, 11] == 1
assert sum(isCC) == full_labels.shape[0], "Have NC events in data"
#Transform Input Data
from scaler_transformations import TransformData, new_transform
full_features_DC = new_transform(full_features_DC)
full_features_DC = TransformData(full_features_DC,
low_stats=low_stat_DC,
high_stats=high_stat_DC,
scaler=transform)
print("Finished DC")
full_features_IC = new_transform(full_features_IC)
full_features_IC = TransformData(full_features_IC,
low_stats=low_stat_IC,
high_stats=high_stat_IC,
scaler=transform)
print("Finished IC")
print("Finished transforming the data using %s Scaler" % transform)
if shuffle:
print("Starting shuffle...")
from handle_data import Shuffler
features_DC, features_IC, labels, \
old_reco, initial_stats, num_pulses = \
Shuffler(features_DC,features_IC,labels, \
old_reco, initial_stats, num_pulses, use_old_reco_flag=use_old_reco)
print("Finished shuffling...")
#Transform Input Data
print("Starting transformation of input features...")
from scaler_transformations import TransformData, new_transform
features_DC_partial_transform = new_transform(features_DC)
del features_DC
features_DC_full_transform = TransformData(features_DC_partial_transform,
low_stats=low_stat_DC,
high_stats=high_stat_DC,
scaler=transform)
del features_DC_partial_transform
print("Finished DC")
features_IC_partial_transform = new_transform(features_IC)
del features_IC
features_IC_full_transform = TransformData(features_IC_partial_transform,
low_stats=low_stat_IC,
high_stats=high_stat_IC,
scaler=transform)
del features_IC_partial_transform