def make_toys(
pdf,
nToys,
data, ## template for dataset/variables
gen_config, ## parameters for <code>pdf.generate</code>
fit_config={}, ## parameters for <code>pdf.fitTo</code>
init_pars={},
more_vars={},
gen_fun=None, ## generator function ( pdf , varset , **config )
fit_fun=None, ## fit function ( pdf , dataset , **config )
accept_fun=None, ## accept function ( fit-result, pdf, dataset )
silent=True,
progress=True):
"""Make `nToys` pseudoexperiments
- Schematically:
>>> for toy in range ( nToys ) :
>>> ... dataset = gen_fun ( pdf , ... , **gen_config )
>>> ... result = fit_fun ( pdf , dataset , **fit_config )
>>> ... if not accept_fun ( result , pdf , dataset ) : continue
>>> .... < collect statistics here >
For each pseudoexperiment:
1. generate dataset using `pdf` with variables specified
in `data` and configuration specified via `gen_config`
for each generation the parameters of `pdf` are reset
for their initial values and valeus from `init_pars`
2. fit generated dataset with `pdf` using configuration
specified via `fit_config`
- pdf PDF to be used for generation and fitting
- nToys number of pseudoexperiments to generate
- data variable list of variables to be used for dataset generation
- gen_config configuration of <code>pdf.generate</code>
- fit_config configuration of <code>pdf.fitTo</code>
- init_pars redefine these parameters for each pseudoexperiment
- more_vars dictionary of functions to define the additional results
- gen_fun generator function
- fit_fun fitting function
- accept_fun accept function
- silent silent toys?
- progress show progress bar?
It returns a dictionary with fit results for the toys and a dictionary of statistics
>>> pdf = ...
... results, stats = make_toys ( pdf , ## PDF to use
... 1000 , ## number of toys
... [ 'mass' ] , ## variables in dataset
... { 'nEvents' : 5000 } , ## configuration of `pdf.generate`
... { 'ncpus' : 2 } , ## configuration of `pdf.fitTo`
... { 'mean' : 0.0 , 'sigma' : 1.0 } ## parameters to use for generation
... )
Derived parameters can be also retrived via <code>more_vars</code> argument:
>>> ratio = lambda res,pdf : res.ratio('x','y')
>>> more_vars = { 'Ratio' : ratio }
>>> r, s = make_toys ( .... , more_vars = more_vars , ... )
- If `gen_fun` is not specified `generate_data` is used
- If `fit_fun` is not specified `make_fit` is used
- If `accept_fun` is not specified `accept_fit` is used
"""
from ostap.core.ostap_types import string_types, integer_types
assert isinstance ( nToys , integer_types ) and 0 < nToys,\
'Invalid "nToys" argument %s/%s' % ( nToys , type ( nToys ) )
assert gen_config and 'nEvents' in gen_config,\
'Number of events per toy must be specified via "gen_config" %s' % gen_config
## 1. generator function?
if gen_fun is None:
if not silent:
logger.info("make_toys: use default ``generate_data'' function!")
gen_fun = generate_data
assert gen_fun and callable(gen_fun), 'Invalid generator function!'
## 2. fitting function?
if fit_fun is None:
if not silent:
logger.info("make_toys: use default ``make_fit'' function!")
fit_fun = make_fit
assert fit_fun and callable(fit_fun), 'Invalid fit function!'
## 3. accept function?
if accept_fun is None:
if not silent:
logger.info("make_toys: use default ``accept_fit'' function!")
accept_fun = accept_fit
assert accept_fun and callable(accept_fun), 'Invalid accept function!'
import ostap.fitting.roofit
import ostap.fitting.dataset
import ostap.fitting.variables
import ostap.fitting.roofitresult
import ostap.fitting.basic
params = pdf.params()
varset = ROOT.RooArgSet()
if isinstance(data, ROOT.RooAbsData): varset = data.varset()
else:
for v in data:
if isinstance(v, ROOT.RooAbsArg):
varset.add(v)
elif isinstance(v, string_types) and v in params:
varset.add(params[v])
else:
raise TypeError('Invalid variable %s/%s' % (v, type(v)))
fix_pars = vars_transform(params)
fix_init = vars_transform(init_pars)
pdf.load_params(params=fix_pars, silent=silent)
pdf.load_params(params=fix_init, silent=silent)
## save all initial parameters (needed for the final statistics)
params = pdf.params()
fix_all = vars_transform(params)
fitcnf = {}
fitcnf.update(fit_config)
if not 'silent' in fitcnf: fitcnf['silent'] = silent
from collections import defaultdict
results = defaultdict(list)
from ostap.core.core import SE, VE
fits = defaultdict(SE) ## fit statuses
covs = defaultdict(SE) ## covariance matrix quality
## run pseudoexperiments
from ostap.utils.progress_bar import progress_bar
for i in progress_bar(range(nToys), silent=not progress):
## 1. reset PDF parameters
pdf.load_params(params=fix_pars, silent=silent)
pdf.load_params(params=init_pars, silent=silent)
## 2. generate dataset!
## dataset = pdf.generate ( varset = varset , **gen_config )
dataset = gen_fun(pdf, varset=varset, **gen_config)
if not silent:
logger.info('Generated dataset #%d\n%s' % (i, dataset))
## 3. fit it!
r = fit_fun(pdf, dataset, **fitcnf)
if not silent:
logger.info('Fit result #%d\n%s' %
(i, r.table(title='Fit result #%d' % i, prefix='# ')))
## fit status
fits[r.status()] += 1
## covariance matrix quality
covs[r.covQual()] += 1
## ok ?
if accept_fun(r, pdf, dataset):
## 4. save results
rpf = r.params(float_only=True)
for p in rpf:
results[p].append(rpf[p][0])
for v in more_vars:
func = more_vars[v]
results[v].append(func(r, pdf))
results['#'].append(len(dataset))
dataset.clear()
del dataset
del r
## make a final statistics
stats = defaultdict(SE)
for par in results:
pars = results[par]
mvar = par in more_vars
if not mvar: a0 = fix_all.get(par, None)
for v in pars:
v0 = float(v)
stats[par] += v0
if not mvar and not a0 is None and isinstance(
v, VE) and 0 < v.error():
stats['pull:%s' % par] += (v0 - a0) / v.error()
for k in fits:
stats['- Status %s' % k] = fits[k]
for k in covs:
stats['- CovQual %s' % k] = covs[k]
if progress or not silent: print_stats(stats, nToys)
return results, stats
def make_toys2(
gen_pdf, ## pdf to generate toys
fit_pdf, ## pdf to fit
nToys, ## number of pseudoexperiments
data, ## template for dataset/variables
gen_config, ## parameters for <code>pdf.generate</code>
fit_config={}, ## parameters for <code>pdf.fitTo</code>
gen_pars={}, ## gen-parameters to reset/use
fit_pars={}, ## fit-parameters to reset/use
more_vars={}, ## additional results to be calculated
gen_fun=None, ## generator function ( pdf , varset , **gen_config )
fit_fun=None, ## fit function ( pdf , dataset , **fit_config )
accept_fun=None, ## accept function ( fit-result, pdf, dataset )
silent=True,
progress=True,
logger=logger,
frequency=1000):
"""Make `ntoys` pseudoexperiments
- Schematically:
>>> for toy in range ( nToys ) :
>>> ... dataset = gen_fun ( gen_pdf , ... , **gen_config )
>>> ... result = fit_fun ( fit_pdf , dataset , **fit_config )
>>> ... if not accept_fun ( result , fit_pdf , dataset ) : continue
>>> .... < collect statistics here >
For each experiment:
1. generate dataset using `pdf` with variables specified
in `data` and configuration specified via `gen_config`
for each generation the parameters of `pdf` are reset
for their initial values and valeus from `init_pars`
2. fit generated dataset with `pdf` using configuration
specified via `fit_config`
- `pdf` : PDF to be used for generation and fitting
- `nToys` : number of pseudoexperiments to generate
- `data` : variable list of variables to be used for dataset generation
- `gen_config` : configuration of <code>pdf.generate</code>
- `fit_config` : configuration of <code>pdf.fitTo</code>
- `gen_pars` : redefine these parameters for generation of each pseudoexperiment
- `fit_pars` : redefine these parameters for fit of each pseudoexperiment
- `silent` : silent toys?
- `progress` : show progress bar?
- `logger` : use this logger
- `frequency` : how often to dump the intermediate results ?
It returns a dictionary with fit results for the toys and a dictionary of statistics
>>> pdf = ...
... results, stats = make_toys ( pdf , ## PDF to use
... 1000 , ## number of toys
... [ 'mass' ] , ## varibales in dataset
... { 'nEvents' : 5000 } , ## configuration of `pdf.generate`
... { 'ncpus' : 2 } , ## configuration of `pdf.fitTo`
... { 'mean' : 0.0 , 'sigma' : 1.0 } ## parameters to use for generation
... )
"""
from ostap.core.ostap_types import string_types, integer_types
assert isinstance ( nToys , integer_types ) and 0 < nToys,\
'Invalid "nToys" argument %s/%s' % ( nToys , type ( nToys ) )
assert gen_config and 'nEvents' in gen_config,\
'Number of events per toy must be specified via "gen_config" %s' % gen_config
## 1. generator function?
if gen_fun is None:
if not silent:
logger.info("make_toys2: use default ``generate_data'' function!")
gen_fun = generate_data
assert gen_fun and callable(gen_fun), 'Invalid generator function!'
## 2. fitting function?
if fit_fun is None:
if not silent:
logger.info("make_toys2: use default ``make_fit'' function!")
fit_fun = make_fit
assert fit_fun and callable(fit_fun), 'Invalid fit function!'
## 3. accept function?
if accept_fun is None:
if not silent:
logger.info("make_toys2: use default ``accept_fit'' function!")
accept_fun = accept_fit
assert accept_fun and callable(accept_fun), 'Invalid accept function!'
if progress and not silent:
assert isinstance ( frequency , integer_types ) and 0 < frequency,\
"make_toys2: invalid ``frequency'' parameter %s" % frequency
import ostap.fitting.roofit
import ostap.fitting.dataset
import ostap.fitting.variables
import ostap.fitting.roofitresult
import ostap.fitting.basic
gparams = gen_pdf.params()
varset = ROOT.RooArgSet()
if isinstance(data, ROOT.RooAbsData): varset = data.varset()
else:
for v in data:
if isinstance(v, ROOT.RooAbsArg):
varset.add(v)
elif isinstance(v, string_types) and v in gparams:
varset.add(gparams[v])
else:
raise TypeError('Invalid variable %s/%s' % (v, type(v)))
## parameters for generation
fix_gen_init = vars_transform(gparams)
fix_gen_pars = vars_transform(gen_pars)
## parameters for fitting
fparams = fit_pdf.params()
fix_fit_init = vars_transform(fparams)
fix_fit_pars = vars_transform(fit_pars)
fitcnf = {}
fitcnf.update(fit_config)
if not 'silent' in fitcnf: fitcnf['silent'] = silent
from collections import defaultdict
results = defaultdict(list)
from ostap.core.core import SE
fits = defaultdict(SE) ## fit statuses
covs = defaultdict(SE) ## covarinace matrix quality
## run pseudoexperiments
from ostap.utils.progress_bar import progress_bar
for i in progress_bar(range(nToys), silent=not progress):
## 1. reset PDF parameters
gen_pdf.load_params(params=fix_gen_init, silent=silent)
gen_pdf.load_params(params=fix_gen_pars, silent=silent)
## 2. generate dataset!
dataset = gen_fun(gen_pdf, varset=varset, **gen_config)
if not silent: logger.info('Generated dataset #%d\n%s' % (i, dataset))
## 3. reset parameters of fit_pdf
fit_pdf.load_params(params=fix_fit_init, silent=silent)
fit_pdf.load_params(params=fix_fit_pars, silent=silent)
## 4. fit it!
r = fit_fun(fit_pdf, dataset, **fitcnf)
## fit status
fits[r.status()] += 1
## covariance matrix quality
covs[r.covQual()] += 1
## ok ?
if accept_fun(r, fit_pdf, dataset):
## 5. save results
rpf = r.params(float_only=True)
for j in rpf:
results[j].append(rpf[j][0])
for v in more_vars:
func = more_vars[v]
results[v].append(func(r, fit_pdf))
results['#'].append(len(dataset))
results['#sumw'].append(dataset.sumVar('1'))
dataset.clear()
del dataset
if progress or not silent:
if 0 < frequency and 1 <= i and 0 == (i + 1) % frequency:
stats = make_stats(results, fits, covs)
print_stats(stats, i + 1, logger=logger)
## make a final statistics
stats = make_stats(results, fits, covs)
if progress or not silent:
print_stats(stats, nToys, logger=logger)
return results, stats
def make_toys2(
gen_pdf, ## pdf to generate toys
fit_pdf, ## pdf to fit
nToys, ## number of pseudoexperiments
data, ## template for dataset/variables
gen_config, ## parameters for <code>pdf.generate</code>
fit_config={}, ## parameters for <code>pdf.fitTo</code>
gen_pars={}, ## gen-parameters to reset/use
fit_pars={}, ## fit-parameters to reset/use
more_vars={}, ## additional results to be calculated
silent=True,
progress=True):
"""Make `ntoys` pseudoexperiments
For each experiment:
1. generate dataset using `pdf` with variables specified
in `data` and configuration specified via `gen_config`
for each generation the parameters of `pdf` are reset
for their initial values and valeus from `init_pars`
2. fit generated dataset with `pdf` using configuration
specified via `fit_config`
- pdf PDF to be used for generation and fitting
- nToys number of pseudoexperiments to generate
- data variable list of variables to be used for dataset generation
- gen_config configuration of <code>pdf.generate</code>
- fit_config configuration of <code>pdf.fitTo</code>
- gen_pars redefine these parameters for generation of each pseudoexperiment
- fit_pars redefine these parameters for fit of each pseudoexperiment
- silent silent toys?
- progress show progress bar?
It returns a dictionary with fit results for the toys
>>> pdf = ...
... results, stats = make_toys ( pdf , ## PDF to use
... 1000 , ## number of toys
... [ 'mass' ] , ## varibales in dataset
... { 'nEvents' : 5000 } , ## configuration of `pdf.generate`
... { 'ncpus' : 2 } , ## configuration of `pdf.fitTo`
... { 'mean' : 0.0 , 'sigma' : 1.0 } ## parameters to use for generation
... )
"""
from ostap.core.ostap_types import string_types, integer_types
assert isinstance ( nToys , integer_types ) and 0 < nToys,\
'Invalid "nToys" argument %s/%s' % ( nToys , type ( nToys ) )
assert gen_config and 'nEvents' in gen_config,\
'Number of events per toy must be specified via "gen_config" %s' % gen_config
import ostap.fitting.roofit
import ostap.fitting.dataset
import ostap.fitting.variables
import ostap.fitting.roofitresult
import ostap.fitting.basic
gparams = gen_pdf.params()
varset = ROOT.RooArgSet()
if isinstance(data, ROOT.RooAbsData): varset = data.varset()
else:
for v in data:
if isinstance(v, ROOT.RooAbsArg):
varset.add(v)
elif isinstance(v, string_types) and v in gparams:
varset.add(gparams[v])
else:
raise TypeError('Invalid variable %s/%s' % (v, type(v)))
## parameters for generation
fix_gen_init = vars_transform(gparams)
fix_gen_pars = vars_transform(gen_pars)
## parameters for fitting
fparams = fit_pdf.params()
fix_fit_init = vars_transform(fparams)
fix_fit_pars = vars_transform(fit_pars)
fitcnf = {}
fitcnf.update(fit_config)
if not 'silent' in fitcnf: fitcnf['silent'] = silent
from collections import defaultdict
results = defaultdict(list)
## run pseudoexperiments
from ostap.utils.progress_bar import progress_bar
for i in progress_bar(range(nToys), silent=not progress):
## 1. reset PDF parameters
gen_pdf.load_params(None, fix_gen_init, silent=silent)
gen_pdf.load_params(None, fix_gen_pars, silent=silent)
## 2. generate dataset!
dataset = gen_pdf.generate(varset=varset, **gen_config)
if not silent:
logger.info('Generated dataset #%d\n%s' % (i, dataset))
## 3. reset parameters of fit_pdf
fit_pdf.load_params(None, fix_fit_init, silent=silent)
fit_pdf.load_params(None, fix_fit_pars, silent=silent)
## 4. fit it!
r, _ = fit_pdf.fitTo(dataset, **fitcnf)
if not silent:
logger.info('Fit result #%d\n%s' %
(i, r.table(title='Fit result #%d' % i, prefix='# ')))
## skip invalid fits
if r.status(): continue
## 5. save results
rpf = r.params(float_only=True)
for i in rpf:
results[i].append(rpf[i][0])
for v in more_vars:
func = more_vars[v]
results[v].append(func(r, fit_pdf))
dataset.clear()
del dataset
## make a final statistics
from ostap.core.core import SE
stats = defaultdict(SE)
for par in results:
pars = results[par]
for v in pars:
v0 = float(v)
stats[par] += v0
if progress or not silent: print_stats(stats, nToys)
return results, stats