def __call__(self, rescale, mu):
"""
$\sum_{i=0}^{n_{bin}} rate - n_i \log(rate)$ with $rate = \mu s + b$
"""
config = self.config
seed = SEED + self.i_cv * 5
train_generator = Generator(seed)
valid_generator = Generator(seed + 1)
classifier = build_model(self.args, self.i_cv)
X_train, y_train, w_train = train_generator.generate(
rescale, config.CALIBRATED.mu, n_samples=config.N_TRAINING_SAMPLES)
classifier.fit(X_train, y_train, w_train)
X, y, w = valid_generator.generate(
rescale, mu, n_samples=config.N_VALIDATION_SAMPLES)
valid_summaries = classifier.compute_summaries(X,
w,
n_bins=self.n_bins)
test_summaries = classifier.compute_summaries(self.X_test,
self.w_test,
n_bins=self.n_bins)
# Compute NLL
EPSILON = 1e-5 # avoid log(0)
rate = valid_summaries + EPSILON
data_nll = np.sum(poisson_nll(test_summaries, rate))
rescale_constraint = gauss_nll(rescale, config.CALIBRATED.rescale,
config.CALIBRATED_ERROR.rescale)
total_nll = data_nll + rescale_constraint
return total_nll
def run(args, i_cv):
logger = logging.getLogger()
print_line()
logger.info('Running iter n°{}'.format(i_cv))
print_line()
directory = os.path.join(DIRECTORY, f'cv_{i_cv}')
os.makedirs(directory, exist_ok=True)
config = Config()
seed = SEED + i_cv * 5
train_generator = Generator(seed)
valid_generator = Generator(seed + 1)
test_generator = Generator(seed + 2)
N_BINS = 10
X_train, y_train, w_train = train_generator.generate(
*config.CALIBRATED, n_samples=config.N_TRAINING_SAMPLES)
compute_summaries = HistogramSummaryComputer(n_bins=N_BINS).fit(X_train)
result_table = [
run_iter(compute_summaries, i_cv, i, test_config, valid_generator,
test_generator, directory)
for i, test_config in enumerate(config.iter_test_config())
]
result_table = pd.DataFrame(result_table)
result_table.to_csv(os.path.join(directory, 'results.csv'))
logger.info('Plot params')
param_names = config.PARAM_NAMES
for name in param_names:
plot_params(name,
result_table,
title='Likelihood fit',
directory=directory)
return result_table
def run(args, i_cv):
logger = logging.getLogger()
print_line()
logger.info('Running iter n°{}'.format(i_cv))
print_line()
# LOAD/GENERATE DATA
logger.info('Set up data generator')
config = Config()
seed = SEED + i_cv * 5
train_generator = GeneratorTorch(seed, cuda=args.cuda)
valid_generator = Generator(seed+1)
test_generator = Generator(seed+2)
# SET MODEL
logger.info('Set up classifier')
model = build_model(args, i_cv)
os.makedirs(model.results_path, exist_ok=True)
flush(logger)
# TRAINING / LOADING
train_or_load_neural_net(model, train_generator, retrain=args.retrain)
# MEASUREMENT
result_row = {'i_cv': i_cv}
results = []
for test_config in config.iter_test_config():
logger.info(f"Running test set : {test_config.TRUE}, {test_config.N_TESTING_SAMPLES} samples")
for threshold in np.linspace(0, 1, 500):
result_row = {'i_cv': i_cv}
result_row['threshold'] = threshold
result_row.update(test_config.TRUE.to_dict(prefix='true_'))
result_row['n_test_samples'] = test_config.N_TESTING_SAMPLES
X, y, w = valid_generator.generate(*config.TRUE, n_samples=config.N_VALIDATION_SAMPLES)
proba = model.predict_proba(X)
decision = proba[:, 1]
selected = decision > threshold
beta = np.sum(y[selected] == 0)
gamma = np.sum(y[selected] == 1)
result_row['beta'] = beta
result_row['gamma'] = gamma
X, y, w = test_generator.generate(*config.TRUE, n_samples=config.N_VALIDATION_SAMPLES)
proba = model.predict_proba(X)
decision = proba[:, 1]
selected = decision > threshold
n_selected = np.sum(selected)
n_selected_bkg = np.sum(y[selected] == 0)
n_selected_sig = np.sum(y[selected] == 1)
result_row['n'] = n_selected
result_row['b'] = n_selected_bkg
result_row['s'] = n_selected_sig
result_row['s_sqrt_n'] = n_selected_sig / np.sqrt(n_selected)
result_row['s_sqrt_b'] = n_selected_sig / np.sqrt(n_selected)
results.append(result_row.copy())
results = pd.DataFrame(results)
print(results)
return results
def run(args, i_cv):
logger = logging.getLogger()
print_line()
logger.info('Running iter n°{}'.format(i_cv))
print_line()
result_row = {'i_cv': i_cv}
# LOAD/GENERATE DATA
logger.info('Set up data generator')
config = Config()
seed = SEED + i_cv * 5
train_generator = Generator(seed)
valid_generator = Generator(seed + 1)
# test_generator = Generator(seed+2)
results = []
for n_train_samples in N_TRAIN_RANGE:
result_row['n_train_samples'] = n_train_samples
# SET MODEL
logger.info('Set up classifier')
model = build_model(args, i_cv)
os.makedirs(model.results_path, exist_ok=True)
flush(logger)
# TRAINING / LOADING
X_train, y_train, w_train = train_generator.generate(
*config.CALIBRATED, n_samples=n_train_samples)
model.fit(X_train, y_train, w_train)
# CHECK TRAINING
logger.info('Generate validation data')
X_valid, y_valid, w_valid = valid_generator.generate(
*config.CALIBRATED, n_samples=config.N_VALIDATION_SAMPLES)
some_eval = evaluate_classifier(model,
X_valid,
y_valid,
w_valid,
prefix='valid',
suffix=f'-{n_train_samples}')
result_row['valid_auc'] = some_eval[f'valid_auc-{n_train_samples}']
result_row['valid_accuracy'] = some_eval[
f'valid_accuracy-{n_train_samples}']
N_BINS = 10
evaluate_summary_computer(model,
X_valid,
y_valid,
w_valid,
n_bins=N_BINS,
prefix='valid_',
suffix=f'{n_train_samples}')
results.append(result_row.copy())
result_table = pd.DataFrame(results)
return result_table
def run(args, i_cv):
logger = logging.getLogger()
print_line()
logger.info('Running iter n°{}'.format(i_cv))
print_line()
result_row = {'i_cv': i_cv}
# LOAD/GENERATE DATA
logger.info('Set up data generator')
config = Config()
seed = SEED + i_cv * 5
train_generator = Generator(seed)
valid_generator = Generator(seed + 1)
test_generator = Generator(seed + 2)
train_generator = TrainGenerator(param_generator, train_generator)
# SET MODEL
logger.info('Set up regressor')
model = build_model(args, i_cv)
os.makedirs(model.results_path, exist_ok=True)
flush(logger)
# TRAINING / LOADING
train_or_load_neural_net(model, train_generator, retrain=args.retrain)
# CHECK TRAINING
logger.info('Generate validation data')
X_valid, y_valid, w_valid = valid_generator.generate(
*config.CALIBRATED, n_samples=config.N_VALIDATION_SAMPLES)
result_row.update(evaluate_neural_net(model, prefix='valid'))
evaluate_regressor(model, prefix='valid')
# MEASUREMENT
result_row['nfcn'] = NCALL
iter_results = [
run_iter(model, result_row, i, test_config, valid_generator,
test_generator)
for i, test_config in enumerate(config.iter_test_config())
]
result_table = [e0 for e0, e1 in iter_results]
result_table = pd.DataFrame(result_table)
result_table.to_csv(os.path.join(model.results_path, 'estimations.csv'))
logger.info('Plot params')
param_names = config.PARAM_NAMES
for name in param_names:
plot_params(name,
result_table,
title=model.full_name,
directory=model.results_path)
conditional_estimate = pd.concat([e1 for e0, e1 in iter_results])
conditional_estimate['i_cv'] = i_cv
fname = os.path.join(model.results_path, "conditional_estimations.csv")
conditional_estimate.to_csv(fname)
logger.info('DONE')
return result_table, conditional_estimate
def main():
# BASIC SETUP
logger = set_logger()
args = REG_parse_args(
main_description="Training launcher for Regressor on S3D2 benchmark")
logger.info(args)
flush(logger)
# Setup model
logger.info("Setup model")
model = build_model(args, 0)
os.makedirs(model.results_directory, exist_ok=True)
# Setup data
logger.info("Setup data")
config = Config()
config_table = evaluate_config(config)
config_table.to_csv(
os.path.join(model.results_directory, 'config_table.csv'))
seed = SEED + 99999
train_generator = TrainGenerator(param_generator, Generator(seed))
valid_generator = Generator(seed + 1)
test_generator = Generator(seed + 2)
i_cv = 0
result_row = {'i_cv': i_cv}
# TRAINING / LOADING
train_or_load_neural_net(model, train_generator, retrain=args.retrain)
# CHECK TRAINING
result_row.update(evaluate_neural_net(model, prefix='valid'))
evaluate_regressor(model, prefix='valid')
print_line()
result_table = [
run_iter(model, result_row, i, test_config, valid_generator,
test_generator)
for i, test_config in enumerate(config.iter_test_config())
]
result_table = pd.DataFrame(result_table)
result_table.to_csv(os.path.join(model.results_directory, 'results.csv'))
logger.info('Plot params')
param_names = [CALIB_PARAM_NAME]
for name in param_names:
plot_params(name,
result_table,
title=model.full_name,
directory=model.results_directory)
logger.info('DONE')
def run(args, i_cv):
logger = logging.getLogger()
print_line()
logger.info('Running iter n°{}'.format(i_cv))
print_line()
# LOAD/GENERATE DATA
logger.info('Set up data generator')
config = Config()
seed = SEED + i_cv * 5
train_generator = Generator(seed)
train_generator = TrainGenerator(param_generator, train_generator)
valid_generator = Generator(seed+1)
test_generator = Generator(seed+2)
# SET MODEL
logger.info('Set up classifier')
model = build_model(args, i_cv)
os.makedirs(model.results_path, exist_ok=True)
flush(logger)
# TRAINING / LOADING
train_or_load_pivot(model, train_generator, config.N_TRAINING_SAMPLES*N_AUGMENT, retrain=args.retrain)
some_fisher = compute_fisher(*compute_bins(model, valid_generator, config, n_bins=3), config.TRUE.mu)
some_fisher_bis = compute_fisher(*compute_bins(model, valid_generator, config, n_bins=3), config.TRUE.mu)
assert some_fisher == some_fisher_bis, f"Fisher info should be deterministic but found : {some_fisher} =/= {some_fisher_bis}"
# MEASUREMENT
result_row = {'i_cv': i_cv}
results = []
for test_config in config.iter_test_config():
logger.info(f"Running test set : {test_config.TRUE}, {test_config.N_TESTING_SAMPLES} samples")
for n_bins in range(1, 30):
result_row = {'i_cv': i_cv}
gamma_array, beta_array = compute_bins(model, valid_generator, test_config, n_bins=n_bins)
fisher = compute_fisher(gamma_array, beta_array, test_config.TRUE.mu)
result_row.update({f'gamma_{i}' : gamma for i, gamma in enumerate(gamma_array, 1)})
result_row.update({f'beta_{i}' : beta for i, beta in enumerate(beta_array, 1)})
result_row.update(test_config.TRUE.to_dict(prefix='true_'))
result_row['n_test_samples'] = test_config.N_TESTING_SAMPLES
result_row['fisher'] = fisher
result_row['n_bins'] = n_bins
results.append(result_row.copy())
results = pd.DataFrame(results)
print(results)
return results
def run_estimation(args, i_cv):
logger = logging.getLogger()
print_line()
logger.info('Running iter n°{}'.format(i_cv))
print_line()
result_row = {'i_cv': i_cv}
# LOAD/GENERATE DATA
logger.info('Set up data generator')
config = Config()
seed = SEED + i_cv * 5
train_generator = Generator(seed)
valid_generator = Generator(seed+1)
test_generator = Generator(seed+2)
# SET MODEL
logger.info('Set up classifier')
model = build_model(args, i_cv)
os.makedirs(model.results_path, exist_ok=True)
flush(logger)
# TRAINING / LOADING
train_or_load_classifier(model, train_generator, config.CALIBRATED, config.N_TRAINING_SAMPLES, retrain=args.retrain)
# CHECK TRAINING
logger.info('Generate validation data')
X_valid, y_valid, w_valid = valid_generator.generate(*config.CALIBRATED, n_samples=config.N_VALIDATION_SAMPLES)
result_row.update(evaluate_classifier(model, X_valid, y_valid, w_valid, prefix='valid'))
# MEASUREMENT
calib_rescale = load_calib_rescale(DATA_NAME, BENCHMARK_NAME)
evaluate_summary_computer(model, X_valid, y_valid, w_valid, n_bins=N_BINS, prefix='valid_', suffix='')
iter_results = [run_estimation_iter(model, result_row, i, test_config, valid_generator, test_generator, calib_rescale, n_bins=N_BINS)
for i, test_config in enumerate(config.iter_test_config())]
result_table = pd.DataFrame(iter_results)
result_table.to_csv(os.path.join(model.results_path, 'estimations.csv'))
logger.info('Plot params')
param_names = config.PARAM_NAMES
for name in param_names:
plot_params(name, result_table, title=model.full_name, directory=model.results_path)
logger.info('DONE')
return result_table
def run_conditional_estimation(args, i_cv):
logger = logging.getLogger()
print_line()
logger.info('Running iter n°{}'.format(i_cv))
print_line()
result_row = {'i_cv': i_cv}
# LOAD/GENERATE DATA
logger.info('Set up data generator')
config = Config()
seed = SEED + i_cv * 5
train_generator = Generator(seed)
valid_generator = Generator(seed+1)
test_generator = Generator(seed+2)
# SET MODEL
logger.info('Set up classifier')
model = build_model(args, i_cv)
os.makedirs(model.results_path, exist_ok=True)
flush(logger)
# TRAINING / LOADING
train_or_load_classifier(model, train_generator, config.CALIBRATED, config.N_TRAINING_SAMPLES, retrain=args.retrain)
# CHECK TRAINING
logger.info('Generate validation data')
X_valid, y_valid, w_valid = valid_generator.generate(*config.CALIBRATED, n_samples=config.N_VALIDATION_SAMPLES)
result_row.update(evaluate_classifier(model, X_valid, y_valid, w_valid, prefix='valid'))
# MEASUREMENT
evaluate_summary_computer(model, X_valid, y_valid, w_valid, n_bins=N_BINS, prefix='valid_', suffix='')
iter_results = [run_conditional_estimation_iter(model, result_row, i, test_config, valid_generator, test_generator, n_bins=N_BINS)
for i, test_config in enumerate(config.iter_test_config())]
conditional_estimate = pd.concat(iter_results)
conditional_estimate['i_cv'] = i_cv
fname = os.path.join(model.results_path, "conditional_estimations.csv")
conditional_estimate.to_csv(fname)
logger.info('DONE')
return conditional_estimate
def explore_distribs():
config = Config()
generator = Generator()
data, label = generator.sample_event(*config.TRUE,
size=config.N_TESTING_SAMPLES)
prior_rescale = stats.norm(loc=config.CALIBRATED.rescale,
scale=config.CALIBRATED_ERROR.rescale)
prior_mu = stats.uniform(loc=0, scale=1)
plot_data_distrib(generator, config)
plot_prior(prior_rescale, "rescale")
plot_prior(prior_mu, "mu")
def run_iter(i_cv, i_iter, config, seed, directory):
logger = logging.getLogger()
result_row = dict(i_cv=i_cv, i=i_iter)
iter_directory = os.path.join(directory, f'iter_{i_iter}')
os.makedirs(iter_directory, exist_ok=True)
logger.info(f"True Parameters = {config.TRUE}")
suffix = f'-mu={config.TRUE.mu:1.2f}_rescale={config.TRUE.rescale}'
generator = Generator(seed) # test_generator
data, label = generator.sample_event(*config.TRUE,
size=config.N_TESTING_SAMPLES)
debug_label(label)
compute_nll = lambda rescale, mu: generator.nll(data, rescale, mu)
plot_nll_around_min(compute_nll, config.TRUE, iter_directory, suffix)
logger.info('Prepare minuit minimizer')
minimizer = get_minimizer(compute_nll, config.CALIBRATED,
config.CALIBRATED_ERROR)
result_row.update(evaluate_minuit(minimizer, config.TRUE))
return result_row
def explore_links():
config = Config()
generator = Generator()
rescale_range = np.linspace(min(config.RANGE.rescale),
max(config.RANGE.rescale),
num=5)
mu_range = np.linspace(min(config.RANGE.mu), max(config.RANGE.mu), num=15)
for rescale in rescale_range:
average_list = []
target_list = []
for mu in mu_range:
data, label = generator.sample_event(rescale,
mu,
size=config.N_TESTING_SAMPLES)
average_list.append(np.mean(data, axis=0))
target_list.append(mu)
plt.scatter(average_list, target_list, label=f'rescale={rescale}')
plt.title('Link between mean(x) and mu')
plt.ylabel('mu')
plt.xlabel('mean(x)')
plt.legend()
plt.savefig(os.path.join(DIRECTORY, 'mean_link.png'))
plt.clf()
def get_generators(i_cv=0):
seed = SEED + i_cv * 5
train_generator = Generator(seed)
valid_generator = Generator(seed + 1)
test_generator = Generator(seed + 2)
return train_generator, valid_generator, test_generator
def run_iter(i_cv, i_iter, config, seed, directory):
# Init
logger = logging.getLogger()
print_line()
logger.info('running iter n°{}'.format(i_iter))
directory = os.path.join(directory, f'iter_{i_iter}')
os.makedirs(directory, exist_ok=True)
results = dict(i_cv=i_cv, i=i_iter)
# Config
RESCALE_MIN = config.TRUE.rescale - 0.2
RESCALE_MAX = config.TRUE.rescale + 0.2
MU_MIN = max(0, config.TRUE.mu - 0.1)
MU_MAX = min(1.0, config.TRUE.mu + 0.1)
MU_N_SAMPLES = 142
RESCALE_N_SAMPLES = 145
DATA_N_SAMPLES = 2000
# Prior
prior_rescale = stats.uniform(loc=RESCALE_MIN,
scale=RESCALE_MAX - RESCALE_MIN)
prior_mu = stats.uniform(loc=MU_MIN, scale=MU_MAX - MU_MIN)
# Param grid
rescale_grid = np.linspace(RESCALE_MIN, RESCALE_MAX, RESCALE_N_SAMPLES)
mu_grid = np.linspace(MU_MIN, MU_MAX, MU_N_SAMPLES)
# Data Generator
generator = Generator(seed)
data, label = generator.sample_event(*config.TRUE, size=DATA_N_SAMPLES)
debug_label(label)
# Compute likelihood
shape = (RESCALE_N_SAMPLES, MU_N_SAMPLES)
n_elements = np.prod(shape)
logger.info(f"3D grid has {n_elements} elements")
log_likelihood = np.zeros(shape)
log_prior_proba = np.zeros(shape)
for i, j in get_iter_prod(RESCALE_N_SAMPLES,
MU_N_SAMPLES,
progress_bar=True):
log_likelihood[i, j] = generator.log_proba_density(
data, rescale_grid[i], mu_grid[j]).sum()
log_prior_proba[i, j] = prior_rescale.logpdf(
rescale_grid[i]) + prior_mu.logpdf(mu_grid[j])
debug_log_proba(log_likelihood, log_prior_proba)
# Normalization
posterior_rescale_mu = softmax(log_likelihood + log_prior_proba)
debug_posterior(posterior_rescale_mu)
# Marginal posterior param proba
marginal_rescale = posterior_rescale_mu.sum(axis=1)
marginal_mu = posterior_rescale_mu.sum(axis=0)
assert marginal_rescale.shape == rescale_grid.shape, "sum along the wrong axis for marginal rescale"
assert marginal_mu.shape == mu_grid.shape, "sum along the wrong axis for marginal mu"
debug_marginal(marginal_rescale, "rescale")
debug_marginal(marginal_mu, "mu")
# Conditional posterior
posterior_mu = np.divide(posterior_rescale_mu,
marginal_rescale.reshape(RESCALE_N_SAMPLES, 1),
out=np.zeros_like(posterior_rescale_mu),
where=(posterior_rescale_mu != 0))
# Minor check
logger.debug("probability densities should sum to one")
debug_proba_sum_one(posterior_mu * marginal_rescale.reshape(-1, 1))
debug_proba_sum_one(posterior_rescale_mu)
debug_proba_sum_one(marginal_rescale)
debug_proba_sum_one(marginal_mu)
# Compute estimator values
sig_ratio = np.sum(label == 1) / DATA_N_SAMPLES
expect_mu = expectancy(mu_grid, marginal_mu)
var_mu = variance(mu_grid, marginal_mu)
std_mu = np.sqrt(var_mu)
expect_rescale = expectancy(rescale_grid, marginal_rescale)
var_rescale = variance(rescale_grid, marginal_rescale)
std_rescale = np.sqrt(var_rescale)
stat_err = stat_uncertainty(mu_grid, posterior_mu, marginal_rescale)
syst_err = syst_uncertainty(mu_grid, posterior_mu, marginal_rescale)
i_max, j_max = np.unravel_index(np.argmax(log_likelihood),
log_likelihood.shape)
assert np.max(log_likelihood) == log_likelihood[
i_max, j_max], "max and argmax should point to the same value"
# Save estimator values
results['mu'] = expect_mu
results['mu' + _TRUTH] = config.TRUE.mu
results['mu_std'] = std_mu
results['mu' + _ERROR] = var_mu
results['mu_stat'] = stat_err
results['mu_syst'] = syst_err
results['rescale'] = expect_rescale
results['rescale' + _TRUTH] = config.TRUE.rescale
results['rescale_std'] = std_rescale
results['rescale' + _ERROR] = var_rescale
# Log estimator values
logger.info(f"True mu value = {config.TRUE.mu}")
logger.info(f"Sig ratio = {sig_ratio}")
logger.info(f"E[mu|x] = {expect_mu}")
logger.info(f"Var[mu|x] = {var_mu}")
logger.info(f"sqrt(Var[mu|x]) = {std_mu}")
logger.info(f"stat_uncertainty = {stat_err}")
logger.info(f"syst_uncertainty = {syst_err}")
logger.info(f"Var - stat = {var_mu - stat_err}")
logger.info(f"argmax_mu p(mu|x) = {mu_grid[np.argmax(marginal_mu)]}")
logger.info(
f"argmax_rescale_mu logp(x|rescale, mu) = {rescale_grid[i_max]} {mu_grid[j_max]}"
)
# Minor checks
debug_min_max(marginal_mu, 'p(mu | x)')
debug_min_max(marginal_rescale, 'p(rescale | x)')
debug_min_max(posterior_mu, 'p(mu | x, rescale)')
debug_min_max(posterior_rescale_mu, 'p(mu, rescale | x)')
# Plots
plot_infer(mu_grid,
marginal_mu,
expected_value=expect_mu,
true_value=config.TRUE.mu,
std=std_mu,
name='mu',
directory=directory,
fname='marginal_mu.png')
plot_infer(rescale_grid,
marginal_rescale,
expected_value=expect_rescale,
true_value=config.TRUE.rescale,
std=std_rescale,
name='rescale',
directory=directory,
fname='marginal_rescale.png')
plot_distrib(data,
generator,
config.TRUE,
expect_rescale,
expect_mu,
title="data distribution",
directory=directory,
fname='data_distrib.png')
return results
def main():
# BASIC SETUP
logger = set_logger()
args = GB_parse_args(
main_description=
"Training launcher for Gradient boosting on S3D2 benchmark")
logger.info(args)
flush(logger)
# Config
config = Config()
config.TRUE = Parameter(rescale=0.9, mu=0.1)
train_generator = Generator(SEED)
valid_generator = Generator(SEED + 1)
test_generator = Generator(SEED + 2)
X_test, y_test, w_test = test_generator.generate(
*config.TRUE, n_samples=config.N_TESTING_SAMPLES)
# for nuisance in p(nuisance | data)
nuisance_param_sample = [
param_generator().nuisance_parameters for _ in range(25)
]
average_list = []
variance_list = []
result_table = []
for nuisance_params in nuisance_param_sample:
logger.info(f"nuisance_params = {nuisance_params}")
estimator_values = []
for i_cv in range(N_ITER):
clf = build_model(args, i_cv)
parameters = Parameter(*nuisance_params,
config.CALIBRATED.interest_parameters)
print(parameters)
n_samples = config.N_TRAINING_SAMPLES
X_train, y_train, w_train = train_generator.generate(
*parameters, n_samples=n_samples)
logger.info(f"Training {clf.full_name}")
clf.fit(X_train, y_train, w_train)
compute_summaries = ClassifierSummaryComputer(clf, n_bins=10)
nll_computer = NLLComputer(compute_summaries,
valid_generator,
X_test,
w_test,
config=config)
compute_nll = lambda mu: nll_computer(*nuisance_params, mu)
minimizer = get_minimizer(compute_nll)
results = evaluate_minuit(minimizer,
[config.TRUE.interest_parameters])
estimator_values.append(results['mu'])
results['i_cv'] = i_cv
results.update(params_to_dict(parameters, suffix='true'))
result_table.append(results.copy())
average_list.append(np.mean(estimator_values))
variance_list.append(np.var(estimator_values))
model = build_model(args, 0)
model.set_info(DATA_NAME, BENCHMARK_NAME, 0)
save_directory = model.results_path
os.makedirs(save_directory, exist_ok=True)
result_table = pd.DataFrame(result_table)
result_table.to_csv(os.path.join(save_directory, 'results.csv'))
logger.info(f"average_list {average_list}")
logger.info(f"variance_list {variance_list}")
v_stat = np.mean(variance_list)
v_syst = np.var(average_list)
v_total = v_stat + v_syst
logger.info(f"V_stat = {v_stat}")
logger.info(f"V_syst = {v_syst}")
logger.info(f"V_total = {v_total}")
eval_dict = {"V_stat": v_stat, "V_syst": v_syst, "V_total": v_total}
eval_path = os.path.join(save_directory, 'info.json')
with open(eval_path, 'w') as f:
json.dump(eval_dict, f)