lattice/pyerrors
1
0
Fork 0
mirror of https://github.com/fjosw/pyerrors.git synced 2025-05-15 03:53:41 +02:00
Code Issues Projects Releases Packages Wiki Activity

fix: pinv replaced by inv for inversion of hessian in fit module, exceptions and warnings added.

Browse source
This commit is contained in:
Fabian Joswig 2022-05-25 14:32:49 +01:00
parent f62581ba69
commit 9ad0146d96
1 changed files with 18 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

20
pyerrors/fits.py
View file

@ -263,10 +263,18 @@ def total_least_squares(x, y, func, silent=False, **kwargs):
output.chisquare_by_expected_chisquare) output.chisquare_by_expected_chisquare)
fitp = out.beta fitp = out.beta
hess = jacobian(jacobian(odr_chisquare))(np.concatenate((fitp, out.xplus.ravel())))
condn = np.linalg.cond(hess)
if condn > 1e8:
warnings.warn("Hessian matrix might be ill-conditioned ({0:1.2e}), error propagation might be unreliable.".format(condn), RuntimeWarning)
try: try:
hess_inv = np.linalg.pinv(jacobian(jacobian(odr_chisquare))(np.concatenate((fitp, out.xplus.ravel())))) hess_inv = np.linalg.inv(hess)
except TypeError: except TypeError:
raise Exception("It is required to use autograd.numpy instead of numpy within fit functions, see the documentation for details.") from None raise Exception("It is required to use autograd.numpy instead of numpy within fit functions, see the documentation for details.") from None
except np.linalg.LinAlgError:
raise Exception("Cannot invert hessian matrix.")
except Exception:
raise Exception("Unkown error in connection with Hessian inverse.")
def odr_chisquare_compact_x(d): def odr_chisquare_compact_x(d):
model = func(d[:n_parms], d[n_parms:n_parms + m].reshape(x_shape)) model = func(d[:n_parms], d[n_parms:n_parms + m].reshape(x_shape))
@ -541,10 +549,18 @@ def _standard_fit(x, y, func, silent=False, **kwargs):
output.chisquare_by_expected_chisquare) output.chisquare_by_expected_chisquare)
fitp = fit_result.x fitp = fit_result.x
hess = jacobian(jacobian(chisqfunc))(fitp)
condn = np.linalg.cond(hess)
if condn > 1e8:
warnings.warn("Hessian matrix might be ill-conditioned ({0:1.2e}), error propagation might be unreliable.".format(condn), RuntimeWarning)
try: try:
hess_inv = np.linalg.pinv(jacobian(jacobian(chisqfunc))(fitp)) hess_inv = np.linalg.inv(hess)
except TypeError: except TypeError:
raise Exception("It is required to use autograd.numpy instead of numpy within fit functions, see the documentation for details.") from None raise Exception("It is required to use autograd.numpy instead of numpy within fit functions, see the documentation for details.") from None
except np.linalg.LinAlgError:
raise Exception("Cannot invert hessian matrix.")
except Exception:
raise Exception("Unkown error in connection with Hessian inverse.")
if kwargs.get('correlated_fit') is True: if kwargs.get('correlated_fit') is True:
def chisqfunc_compact(d): def chisqfunc_compact(d):