def test_const_results(self):
pdf = PDF('NNPDF30_nlo_as_0118')
obs = make_observable('data/higgs/ggh_13tev.root', order=1)
res = produce_results(pdf, obs)[0]._data.as_matrix()
previous = [ 31.22150023, 28.9888971 , 31.3305855 , 31.35700134,
30.66074282, 28.2795592 , 31.50932389, 31.43558106,
31.28601552, 31.36406676, 30.69818304, 30.86453794,
31.62708945, 31.6582997 , 31.11008626, 31.59319744,
31.52866286, 31.28639456, 31.07427256, 31.41568155,
31.60044526, 31.36966168, 30.93032191, 33.05392731,
30.96536474, 31.4317893 , 30.72436163, 31.22605137,
31.24243901, 31.21880536, 31.68798426, 31.07646885,
30.59035729, 31.34800315, 31.21256586, 32.19365014,
31.58647388, 31.74649083, 31.04692704, 30.2313781 ,
31.4132261 , 31.01493873, 31.36481456, 31.26872855,
31.09018502, 31.51168931, 31.28105268, 30.91102961,
30.88952926, 31.31389055, 31.22196011, 31.19094082,
31.03705732, 31.54684624, 31.25938782, 31.10403155,
30.87467055, 31.83239232, 31.05799506, 31.68930494,
31.0387522 , 31.27327964, 31.30919955, 31.51614659,
31.06029472, 31.43211917, 31.35963882, 32.68355944,
30.96018759, 30.60001451, 31.25165147, 31.37222175,
31.19241142, 29.4805518 , 31.66853784, 30.96975209,
30.61955437, 31.34032813, 32.0724701 , 30.9346566 ,
31.27130234, 31.38000397, 30.83068422, 31.50593514,
31.22489125, 30.74042979, 31.21191289, 31.4358684 ,
31.27231758, 31.81871636, 31.10093208, 30.64649533,
31.40680264, 31.12985161, 31.8378625 , 31.18650963,
31.51129514, 31.18025208, 31.87366192, 31.78494394, 31.45846653]
self.assertTrue(np.allclose(res, previous))
def execute_config(conf, output_dir, db):
import pandas as pd
import smpdflib.core as lib
import logging
resultset = []
for group in conf.actiongroups:
pdfsets, observables = group['pdfsets'], group['observables']
resources = group.copy()
resources.pop('actions')
# perform convolution
#TODO Do this better
if any(requires_result(act) for act in group['actions']):
results = lib.produce_results(pdfsets, observables, db)
resultset.append(results)
data_table = lib.results_table(results)
summed_table = lib.summed_results_table(results)
total = pd.concat((data_table,
summed_table),
ignore_index = True)
if logging.getLogger().isEnabledFor(logging.DEBUG):
...
#print_results(results)
resources.update({'results':results, 'data_table':data_table,
'total':total, 'summed_table':summed_table})
if any(requires_correlations(act) for act in group['actions']):
pdfcorrlist = lib.correlations(data_table, db=db)
resources.update({'pdfcorrlist':pdfcorrlist})
prefix = group['prefix']
resources.update({ 'output_dir':output_dir,
'prefix':prefix,
'pdfsets': pdfsets,
'db': db})
for action, res in do_actions(group['actions'], resources):
logging.info("Finalized action '%s'." % action)
return resultset
pdf = PDF("MC900_nlo")
obs = [make_observable(path, order='NLO') for path in itertools.chain(
sorted(glob.glob("data/z/*.root")),
sorted(glob.glob("data/w/*.root")),
sorted(glob.glob("data/higgs/*.root")),
sorted(glob.glob("data/ttbar/*.root")),
)]
tolerance = 0.05
thresholds = [0.0, 0.25, 0.5, 0.75, 0.9, 0.99]
if __name__ == '__main__':
neig = []
logging.basicConfig(level=logging.INFO)
db = shelve.open('db/db')
results = produce_results([pdf], obs, db=db)
for t in thresholds:
V, _ ,desc = get_smpdf_params(pdf, results, smpdf_tolerance=tolerance,
correlation_threshold=t, db=db)
neig.append(V.shape[1])
print("For thresholf %.2f we get:" %t)
print(desc)
with open("thresholdsladder.json", 'w') as f:
json.dump([thresholds, neig], f)
db.close()
def get_smpdf_params(
pdf,
pdf_results,
smpdf_tolerance,
full_grid=False,
db=None,
correlation_threshold=DEFAULT_CORRELATION_THRESHOLD,
nonlinear_correction=True,
):
first_res = get_smpdf_lincomb(
pdf, pdf_results, full_grid=full_grid, target_error=smpdf_tolerance, correlation_threshold=correlation_threshold
)
norm = first_res.norm
lincomb = first_res.lincomb
description = first_res.desc
vec = first_res.lincomb / norm
if nonlinear_correction:
logging.info("Estimating nonlinear correction")
with tempfile.TemporaryDirectory() as td:
applwrap.setlhapdfpath(td)
tempname = str(uuid.uuid1())
logging.info("Creating temporary PDF %s" % tempname)
hessian_from_lincomb(pdf, vec, folder=td, set_name=tempname, db=db)
observables = [r.obs for r in pdf_results]
temppdf = PDF(tempname)
temppdf.infopath
real_results = produce_results(temppdf, observables)
logging.info("Real results obtained")
results_to_refine = []
newtols = []
for smpdf_res, prior_res in zip(real_results, pdf_results):
real_error = 1 - smpdf_res.std_error() / prior_res.std_error()
# pandas indexing is broken, so have to call as_matrix....
bad_bins = np.array(real_error > smpdf_tolerance, copy=False)
if bad_bins.any():
lin_errors = list(first_res.errors[str(smpdf_res.obs)].values())
newtol = smpdf_tolerance - (real_error[bad_bins] - np.array(lin_errors)[bad_bins])
impossible = np.argwhere(newtol < 0)
if len(impossible):
raise TooMuchPrecision(smpdf_res.obs, impossible[0] + 1)
newtols += list(newtol)
logging.debug("New tolerances for observable %s: %s" % (prior_res.obs, newtol))
# Create result with the same type as prior, and only
# bad_bins.
newres = type(prior_res)(prior_res.obs, prior_res.pdf, prior_res._data.ix[bad_bins])
results_to_refine.append(newres)
if results_to_refine:
logging.info("Calculating eigenvectors to refine")
ref_res = get_smpdf_lincomb(
pdf,
results_to_refine,
full_grid=full_grid,
target_error=newtols,
correlation_threshold=correlation_threshold,
Rold=first_res.Rold,
)
lincomb, description = merge_lincombs(first_res.lincomb, ref_res.lincomb, description, ref_res.desc)
else:
logging.info("All results are within tolerance")
return lincomb, norm, description
from smpdflib.plots import plot_bindist
#Db folder should exist...
db = shelve.open('db/db')
to_plot = {make_observable('data/applgrid/CMSWCHARM-WpCb-eta4.root', order=1) : 4,
make_observable('data/applgrid/APPLgrid-LHCb-Z0-ee_arXiv1212.4260-eta34.root', order=1) : 8,
}
pdfs = [PDF('MC900_nnlo', label="legendtext"),
PDF('CMC100_nnlo', label="other"),
PDF('MCH_nnlo_100', label="another")
]
if __name__ == '__main__':
for obs, bin in to_plot.items():
results = produce_results(pdfs, [obs], db=db)
obs_table = results_table(results)
for (obs,b), fig in plot_bindist(obs_table, bin, base_pdf = pdfs[0]):
ax = fig.axes[0]
ax.set_xlabel("My X label")
ax.set_ylabel("MY Y label")
ax.set_title("My title")
path = "%s_bin_%s.pdf" % (obs, b+1)
fig.savefig(path)
plt.close(fig)
prior_name = "MC900_nnlo"
prefixes = ['z2tr' + x for x in '0 10 25 50 75 90 99'.split()]
smpdf_names = [prefix + 'smpdf_' + prior_name for prefix in prefixes]
lincoef_paths = ['output/%s_lincomb.csv' % name for name in smpdf_names]
if __name__ == '__main__':
prior = PDF(prior_name)
smpdfs = [PDF(name) for name in smpdf_names]
pdfs = [prior] + smpdfs
obs = make_observable(obs_name, order='NLO')
with shelve.open('db/db') as db:
res_prior, *res_smpdfs = produce_results(pdfs, [obs], db=db)
coefs = [pd.DataFrame.from_csv(path, sep='\t') for path in lincoef_paths]
prior_std = res_prior.std_error()
real_tols = [1 - res.std_error()/prior_std for res in res_smpdfs]
prior_diffs = (res_prior._all_vals.T - res_prior.central_value).as_matrix().ravel()
neig = [len(pdf) - 1 for pdf in smpdfs]
rotated_tols = []
for path in lincoef_paths:
coefs = pd.DataFrame.from_csv(path, sep='\t')
rotated_diffs = np.dot(prior_diffs, coefs)
rotated_std = la.norm(rotated_diffs)
