def prep(n, labelcol):
"""
Load the data
"""
f = pf.open(n)
data = f[1].data
names = f[1].columns.names
f.close()
try:
labels = data.field(labelcol)
except:
labels = np.zeros(data.field(0).size) - 99
ylim = np.inf
featurenames = ['cmodelmag', 'psffwhm', 'petror50', 'petror90']
targetnames = ['psfmag', 'cmodelmag']
filters = ['u', 'g', 'r', 'i', 'z']
x = FeatureExtractor(data, featurenames, filters, color_band='r', scale_kind=None,
mag_range=None)
data = data[x.idx]
labels = labels[x.idx]
y = FeatureExtractor(data, targetnames, filters, color_band=None, scale_kind=None,
mag_range=None)
# taylor to target, set for psf - model
y.features[:, :5] = y.features[:, :5] - y.features[:, 5:10]
y.features[:, 5:10] = np.sqrt(y.features[:, 10:15] ** 2. + y.features[:, 15:20] ** 2.)
y.features = y.features[:, :10]
# restrict y range
ylim = 10.
ind = y.features[:, 2] < ylim
x.features = x.features[ind]
y.features = y.features[ind]
labels = labels[ind]
y.Ndata = y.features.shape[0]
return x, y, labels
x = FeatureExtractor(data, featurenames, filters, color_band="r", scale_kind=None, mag_range=None)
data = data[x.idx]
y = FeatureExtractor(data, targetnames, filters, color_band=None, scale_kind=None, mag_range=None)
# taylor to target, set for psf - model
y.features[:, :5] = y.features[:, :5] - y.features[:, 5:10]
y.features[:, 5:10] = np.sqrt(y.features[:, 10:15] ** 2.0 + y.features[:, 15:20] ** 2.0)
y.features = y.features[:, :10]
# restrict x range
xlim = (19.5, 20.5)
ind = (x.features[:, 2] > xlim[0]) & (x.features[:, 2] < xlim[1])
x.features = x.features[ind]
y.features = y.features[ind]
y.Ndata = y.features.shape[0]
# restrict y range
ylim = 0.5
ind = y.features[:, 2] < ylim
x.features = x.features[ind]
y.features = y.features[ind]
y.Ndata = y.features.shape[0]
# specify scikit regressor
rname = "RF"
if rname == "KNN":
rgr = KNeighborsRegressor(n_neighbors=8)
if rname == "RF":
rgr = RandomForestRegressor(n_estimators=128)