def go(self, guess):
"""
Run leastsq and set the result
"""
guess = np.array(guess, dtype='f8', copy=False)
nobj = guess.size // self.npars_per
nleft = guess.size % self.npars_per
if nobj != self.nobj or nleft != 0:
raise ValueError("bad guess size: %d" % guess.size)
bounds = self._get_bounds(nobj)
result = run_leastsq(self._calc_fdiff,
guess,
self.n_prior_pars,
k_space=True,
bounds=bounds,
**self.lm_pars)
result['model'] = self.model
if result['flags'] == 0:
stat_dict = self.get_fit_stats(result['pars'])
result.update(stat_dict)
self._result = result
def fit_exp_binned(x, y, yerr, guess):
# assume quadratic
def loss(pars):
model = exp_func(pars, x)
return (model - y) / yerr
return run_leastsq(
loss,
np.array(guess),
0,
)
def fit_exp(x, y, guess):
def loss(pars):
model = exp_func(pars, x)
return (model - y)
return run_leastsq(
loss,
np.array(guess),
0,
maxfev=4000,
xtol=1.0e-5,
ftol=1.0e-5,
)
def fit_erf(x, y, guess, type='falling'):
def func(pars):
return erf_func(pars, x, type)
def loss(pars):
model = func(pars)
return (model - y)
return run_leastsq(
loss,
np.array(guess),
0,
)
def go(self, ntry=100):
ngauss = self.ngauss
# ysum = self.y.sum()
npars_per = 3
for i in range(ntry):
self.guess = np.zeros(npars_per * ngauss)
for ip, prior in enumerate(self.priors):
self.guess[ip] = prior.sample(sigma_factor=1)
# self.guess[ip] = prior.sample(sigma_factor=0.1)
print_pars(self.guess, front='guess: ')
self.result = run_leastsq(
self._errfunc,
self.guess,
self.n_prior_pars,
bounds=self.bounds,
maxfev=4000,
)
if self.result['flags'] == 0:
break