def best_fit(self, **kwargs):
r"""Compute the best fit point in the space of fit parameters and Wilson
coefficients.
Keyword arguments will be passed to `scipy.optimize.minimize_scalar` in
the case of a single fit variable and to `flavio.math.optimize.minimize_robust`
in the case of multiple fit variables.
Returns a dictionary with the following keys:
- 'x': position of the best fit point
- 'log_likelihood': logarithm of the likelihood at the best fit point
"""
n_fit_p = len(self.fit_parameters)
n_wc = len(self.fit_wc_names)
if n_fit_p + n_wc == 1:
def f(x):
return -self.log_likelihood([x])
opt = scipy.optimize.minimize_scalar(f, **kwargs)
else:
def f(x):
return -self.log_likelihood(x)
if 'x0' not in kwargs:
x0 = np.zeros(n_fit_p + n_wc)
if n_fit_p > 1:
x0[:n_fit_p] = self.get_central_fit_parameters
opt = minimize_robust(f, x0, **kwargs)
else:
opt = minimize_robust(f, **kwargs)
if not opt.success:
raise ValueError("Optimization failed.")
else:
return {'x': opt.x, 'log_likelihood': -opt.fun}
def _best_fit_point(initial_guess, gl, wc_fct, scale_fct, array_input,
methods):
llh_name, x0 = initial_guess
llh = (gl.fast_likelihoods.get(llh_name) or gl.likelihoods.get(llh_name)
or gl.custom_likelihoods.get(llh_name)
or (_global_llh(gl) if llh_name == 'global' else None))
llh_sm = (0 if llh_name == 'global' else gl.log_likelihood_sm[llh_name])
if array_input:
def chi2(x):
w = gl.get_wilson(wc_fct(x), scale_fct(x))
gp = gl.parameter_point(w)
try:
res = -2 * (llh.log_likelihood(gp.par_dict_np, w, delta=True) -
llh_sm)
except ValueError as e:
if ('The extraction of CKM elements failed.' in str(e)
or 'math domain error' in str(e)):
res = np.inf
else:
raise
return res
else:
def chi2(x):
w = gl.get_wilson(wc_fct(*x), scale_fct(*x))
gp = gl.parameter_point(w)
try:
res = -2 * (llh.log_likelihood(gp.par_dict_np, w, delta=True) -
llh_sm)
except ValueError as e:
if ('The extraction of CKM elements failed.' in str(e)
or 'math domain error' in str(e)):
res = np.inf
else:
raise
return res
try:
res = minimize_robust(chi2, x0, methods=methods)
if not res.success:
x = [np.nan] * len(x0)
z = np.nan
warnings.warn(
"Optimization failed during computation of best-fit point of {}: {}"
''.format(llh_name, res.message))
else:
x = res.x
z = res.fun
except Exception as e:
x = [np.nan] * len(x0)
z = np.nan
warnings.warn('\nERROR during computation of best-fit point of {}:\n{}'
''.format(llh_name, e))
return (llh_name, {'coords_min': np.array(x), 'z_min': z})
def best_fit(self, **kwargs):
r"""Compute the best fit point in the space of fit parameters and Wilson
coefficients.
Keyword arguments will be passed to `scipy.optimize.minimize_scalar` in
the case of a single fit variable and to `flavio.math.optimize.minimize_robust`
in the case of multiple fit variables.
Returns a dictionary with the following keys:
- 'x': position of the best fit point
- 'log_likelihood': logarithm of the likelihood at the best fit point
"""
n_fit_p = len(self.fit_parameters)
n_wc = len(self.fit_wc_names)
if n_fit_p + n_wc == 1:
def f(x):
return -self.log_likelihood([x])
opt = scipy.optimize.minimize_scalar(f, **kwargs)
else:
def f(x):
return -self.log_likelihood(x)
if 'x0' not in kwargs:
x0 = np.zeros(n_fit_p + n_wc)
if n_fit_p > 1:
x0[:n_fit_p] = self.get_central_fit_parameters
opt = minimize_robust(f, x0, **kwargs)
else:
opt = minimize_robust(f, **kwargs)
if not opt.success:
raise ValueError("Optimization failed.")
else:
return {'x': opt.x, 'log_likelihood': -opt.fun}