def main():
numpy.set_printoptions(precision=2, linewidth=400, threshold=100, sign=' ')
var_edges = numpy.linspace(200, 1200, 31)
kv_val = 1.0
#num_bins = 21
#k2v_val_range = numpy.linspace(-2,4,num_bins)
#kl_val_range = numpy.linspace(-14,16,num_bins)
truth_data_files = fileio_utils.read_coupling_file(coupling_file='basis_files/truth_LHE_couplings.dat')
truth_weights, truth_errors = fileio_utils.extract_lhe_truth_data(truth_data_files.values(), var_edges, stat_limit=None, emulateSelection=True)
def single_negative_weight_map(basis_parameters, name_suffix='_base', truth_level=False):
var_edges = numpy.linspace(200, 1200, 31)
kv_val = 1.0
num_kappa_bins = 100
k2v_val_range = numpy.linspace(-2,4,num_kappa_bins+1)
kl_val_range = numpy.linspace(-14,16,num_kappa_bins+1)
if truth_level:
truth_data_files = fileio_utils.read_coupling_file(coupling_file='basis_files/truth_LHE_couplings.dat')
basis_files = [ truth_data_files[coupling] for coupling in basis_parameters ]
base_weights, base_errors = fileio_utils.extract_lhe_truth_data(basis_files, var_edges)
else:
data_files = fileio_utils.read_coupling_file()
base_events_list = fileio_utils.get_events(basis_parameters, data_files)
base_histograms = [ fileio_utils.retrieve_reco_weights(var_edges, base_events) for base_events in base_events_list ]
base_weights, base_errors = numpy.array(list(zip(*base_histograms)))
integral = get_Nweight_sum(basis_parameters, base_weights, kv_val, k2v_val_range, kl_val_range, grid=False)
negative_weight_grid = get_Nweight_sum(basis_parameters, base_weights, kv_val, k2v_val_range, kl_val_range, grid=True)
draw_error_map(basis_parameters, var_edges, kv_val, k2v_val_range, kl_val_range, negative_weight_grid, name_suffix=name_suffix,
title_suffix=f'Integral={int(integral)}')
def generate_metric_values():
var_edges = numpy.linspace(200, 1200, 31)
kv_val = 1.0
num_bins = 100
k2v_val_range = numpy.linspace(-2, 4, num_bins + 1)
kl_val_range = numpy.linspace(-14, 16, num_bins + 1)
#data_files = fileio_utils.read_coupling_file()
#all_events = fileio_utils.get_events(data_files.keys(), data_files)
#all_histograms = [ fileio_utils.retrieve_reco_weights(var_edges,events) for events in all_events ]
#all_variations = list(zip(data_files.keys(), all_histograms, all_events))#[:7]
#all_cutflows = fileio_utils.get_combined_cutflow_values(data_files.keys(), data_files).values() # It's a really good things that python dicts are ordered...
truth_data_files = fileio_utils.read_coupling_file(
coupling_file='basis_files/truth_LHE_couplings_extended.dat')
truth_weights, truth_errors = fileio_utils.extract_lhe_truth_data(
truth_data_files.values(), var_edges)
all_variations = list(
zip(truth_data_files.keys(), truth_weights, truth_errors))
#assert list(data_files.keys()) == list(truth_data_files.keys())
#all_variations = list(zip(data_files.keys(), all_histograms, all_events, all_events, truth_weights, truth_errors))
# Wrap all variations up together with their histograms so I can find combinations
#all_variations = list(zip(data_files.keys(), all_histograms, all_cutflows, all_events))#[:7]
print('All variations loaded, proceeding to retrieve metrics...')
#for variation, cuts in zip(data_files.keys(), all_cutflows):
# accXeff = cuts['Signal'] / cuts['Initial']
# print(variation, accXeff)
#exit()
total = 0
basis_metrics = {}
for basis_set in itertools.combinations(all_variations, 6):
# Unwrap each combination
#couplings, histograms, cutflows, events_list, truth_weights, truth_errors = list(zip(*basis_set))
#couplings, histograms, events_list = list(zip(*basis_set))
couplings, truth_weights, truth_errors = list(zip(*basis_set))
if not is_valid_combination(couplings): continue
#if (1.0,1.0,1.0) not in couplings: continue
#weights, errors = numpy.array( list(zip(*histograms)) )
basis_metrics[couplings] = {
#'Nweight_integral': get_Nweight_sum(couplings, weights, kv_val, k2v_val_range, kl_val_range),
'Nweight_truth_integral':
get_Nweight_sum(couplings, truth_weights, kv_val, k2v_val_range,
kl_val_range),
#'orthogonality': metric_orthogonality(couplings),
#'reco_effective_stats_integral': metric_reco_effective_stats_integral(couplings, events_list, kv_val, k2v_val_range, kl_val_range),
'theory_effective_stats_integral':
get_theory_effective_stats_map(couplings, kv_val, k2v_val_range,
kl_val_range),
#'reco_solidarity_integral': get_reco_solidarity_map(couplings, weights, kv_val, k2v_val_range, kl_val_range),
'theory_solidarity_integral':
get_theoretical_solidarity_map(couplings, kv_val, k2v_val_range,
kl_val_range),
#'theory_test_val': metric_theory_test_val(couplings),
#'contribution_integral': metric_contribution_integral(couplings, kv_val, k2v_val_range, kl_val_range),
#'accXeff_list': metric_accXeff_list(cutflows),
#'accXeff_sum': metric_accXeff_sum(cutflows),
#'accXeff_geometric': metric_accXeff_geometric_mean(cutflows),
#'accXeff_rms': metric_accXeff_rms(cutflows),
#'accXeff_avg_stdev': metric_accXeff_avg_stdev(cutflows),
#'accXeff_min': metric_accXeff_min(cutflows),
#'accXeff_sigma': metric_accXeff_sigma(cutflows),
#'accXeff_harmonic': metric_accXeff_harmonic_mean(cutflows),
#'eventCount_sum': metric_eventCount_sum(events_list)
}
total += 1
if total % 10 == 0: print(total)
print('Variations traversed, plotting...')
metric_lists = {key: [] for key in list(basis_metrics.values())[0].keys()}
for basis, metrics in basis_metrics.items():
for key, val in metrics.items():
metric_lists[key].append(val)
return metric_lists
def optimize_reco(mode='reco', extra_files={}, extra_name=''):
var_edges = numpy.linspace(200, 1200, 31)
kv_val = 1.0
num_kappa_bins = 100
k2v_val_range = numpy.linspace(-2, 4, num_kappa_bins + 1)
kl_val_range = numpy.linspace(-14, 16, num_kappa_bins + 1)
grid_pixel_area = (k2v_val_range[1] -
k2v_val_range[0]) * (kl_val_range[1] - kl_val_range[0])
truth_data_files = None
if mode == 'reco':
data_files = fileio_utils.read_coupling_file()
all_events = fileio_utils.get_events(data_files.keys(), data_files)
all_histograms = [
fileio_utils.retrieve_reco_weights(var_edges, events)
for events in all_events
]
all_weights, all_errors = numpy.array(list(zip(*all_histograms)))
# Wrap all variations up together with their histograms so I can find combinations
all_variations = list(zip(data_files.keys(), all_weights))
elif mode == 'truth':
truth_data_files = fileio_utils.read_coupling_file(
coupling_file='basis_files/truth_LHE_couplings.dat')
truth_data_files.update(extra_files)
truth_weights, truth_errors = fileio_utils.extract_lhe_truth_data(
truth_data_files.values(), var_edges)
all_variations = list(zip(truth_data_files.keys(), truth_weights))
else:
print('What are you doing?!')
print(mode)
exit(1)
print('Histograms loaded, proceeding to integrate Nweight grids...')
valid_bases = []
total = 0
for basis_set in itertools.combinations(all_variations, 6):
# Unwrap each combination
couplings, weights = list(zip(*basis_set))
if (1.0, 1.0, 1.0) not in couplings: continue
if not combination_utils.is_valid_combination(
couplings, base_equations=combination_utils.full_scan_terms):
continue
nWeight_integral = get_Nweight_sum(couplings, weights, kv_val,
k2v_val_range, kl_val_range)
#nWeight_integral = get_Nweight_sum(couplings, weights, kv_val, k2v_val_range, kl_val_range,
# mask=lambda k2v, kl: ((k2v-1)/1)**2 + ((kl-1)/10)**2 < 1 )
valid_bases.append((nWeight_integral, couplings, weights))
total += 1
if total % 10 == 0: print(total)
print('Integrals computed, sorting and printing...')
valid_bases.sort()
for rank, (integral, couplings, weight) in enumerate(valid_bases):
print(rank, int(integral), couplings)
#draw_rankings([0,1,2,3], valid_bases, var_edges, kv_val, k2v_val_range, kl_val_range, 'quad', only_heatmap=True)
#draw_rankings([0,1], valid_bases, var_edges, kv_val, k2v_val_range, kl_val_range, 'top')
#draw_rankings([0,1,2,3], valid_bases, var_edges, kv_val, k2v_val_range, kl_val_range, extra_name+'_truth_quad', only_heatmap=True)
draw_rankings([0, 1],
valid_bases,
var_edges,
kv_val,
k2v_val_range,
kl_val_range,
extra_name + '_truth_top',
only_heatmap=False,
truth_level=True,
truth_data_files=truth_data_files,
skip_preview=True)
def compare_bases_reco_method(basis_parameters_list,
verification_parameters,
base_equations=combination_utils.full_scan_terms,
name_suffix='',
title_suffix='',
labels=('', ''),
is_verification=True,
truth_level=False,
truth_data_files=None):
#var_edges = numpy.linspace(200, 1200, 31)
#var_edges = numpy.arange(0, 2050, 50)
var_edges = numpy.linspace(200, 2000, 55)
basis_tuple_list = []
for basis_parameters in basis_parameters_list:
reweight_vector = get_amplitude_function(basis_parameters,
as_scalar=False,
base_equations=base_equations)
if truth_level:
data_files = fileio_utils.read_coupling_file(
coupling_file='basis_files/truth_LHE_couplings_extended.dat')
basis_files = [
truth_data_files[coupling] for coupling in basis_parameters
]
truth_weights, truth_errors = fileio_utils.extract_lhe_truth_data(
basis_files, var_edges)
basis_tuple_list.append(
(truth_weights, truth_errors, reweight_vector))
else:
data_files = fileio_utils.read_coupling_file()
base_events_list = fileio_utils.get_events(basis_parameters,
data_files)
base_histograms = [
fileio_utils.retrieve_reco_weights(var_edges, base_events)
for base_events in base_events_list
]
base_weights, base_errors = numpy.array(list(
zip(*base_histograms)))
basis_tuple_list.append(
(base_weights, base_errors, reweight_vector))
testpoint_list = verification_parameters
if is_verification:
if truth_level:
verification_files = [
data_files[key] for key in verification_parameters
]
truth_verification_weights, truth_verification_errors = fileio_utils.extract_lhe_truth_data(
verification_files, var_edges)
testpoint_list = zip(verification_parameters,
truth_verification_weights,
truth_verification_errors)
else:
testpoint_list = []
verification_events_list = fileio_utils.get_events(
verification_parameters, data_files)
for events, param in zip(verification_events_list,
verification_parameters):
verification_weights, verification_errors = fileio_utils.retrieve_reco_weights(
var_edges, events)
testpoint_list.append(
(param, verification_weights, verification_errors))
for testpoint in testpoint_list:
verification_weights, verification_errors = None, None
if is_verification:
coupling_parameters, verification_weights, verification_errors = testpoint
else:
coupling_parameters = testpoint
combined_tuples = []
for base_weights, base_errors, reweight_vector in basis_tuple_list:
combined_tuples.append(
reco_reweight(reweight_vector, coupling_parameters,
base_weights, base_errors))
if truth_level:
name = 'truth_mHH_compare' + name_suffix
title = 'Truth LHE-Based Linear Combination:\nTruth $m_{HH}$' + title_suffix
xlabel = 'Truth $m_{HH}$ (GeV)'
else:
name = 'reco_mHH_compare' + name_suffix
title = 'NNT-Based Linear Combination:\n$m_{HH}$' + title_suffix
xlabel = 'Reconstructed $m_{HH}$ (GeV)'
plot_histogram(
name,
title,
var_edges,
coupling_parameters,
combined_tuples[0][0],
combined_tuples[0][1],
verification_weights,
verification_errors,
alt_linearly_combined_weights=combined_tuples[1][0],
alt_linearly_combined_errors=combined_tuples[1][1],
generated_label=labels[0],
alt_label=labels[1],
xlabel=xlabel,
)