def test_linfit_tied():
p0, parinfo, fa = generate_toy_model()
# This is a bad tie, but its to test it works
parinfo[0]["tied"] = "2 * p[1]"
m = mpfit(myfunctlin, p0, parinfo=parinfo, functkw=fa)
res = m.optimize_result
assert res.success
assert res.dof == 9
assert res.status == 1
np.testing.assert_allclose(m.params[0] * 0.5, m.params[1], rtol=5e-7)
np.testing.assert_allclose(m.params,
np.array([1.60881708, 0.80440854]),
rtol=5e-7)
np.testing.assert_allclose(m.perror,
np.array([0.0, 0.00990717]),
rtol=5e-7)
chisq = np.sum(
myfunctlin(m.params, x=fa["x"], y=fa["y"], err=fa["err"])[1]**2)
np.testing.assert_allclose(np.array([chisq]),
np.array([25465.436783]),
rtol=5e-7)
def test_linfit():
x = N.array([-1.7237128E+00, 1.8712276E+00, -9.6608055E-01,
-2.8394297E-01, 1.3416969E+00, 1.3757038E+00,
-1.3703436E+00, 4.2581975E-02, -1.4970151E-01,
8.2065094E-01])
y = N.array([1.9000429E-01, 6.5807428E+00, 1.4582725E+00,
2.7270851E+00, 5.5969253E+00, 5.6249280E+00,
0.787615, 3.2599759E+00, 2.9771762E+00,
4.5936475E+00])
ey = 0.07 * N.ones(y.shape, dtype='float64')
p0 = N.array([1.0, 1.0], dtype='float64') # initial conditions
pactual = N.array([3.2, 1.78]) # actual values used to make data
parbase = {'value': 0., 'fixed': 0, 'limited': [0, 0], 'limits': [0., 0.]}
parinfo = []
for i in range(len(pactual)):
parinfo.append(copy.deepcopy(parbase))
for i in range(len(pactual)):
parinfo[i]['value'] = p0[i]
fa = {'x': x, 'y': y, 'err': ey}
m = mpfit(myfunctlin, p0, parinfo=parinfo, functkw=fa)
if m.status <= 0:
print 'error message = ', m.errmsg
assert N.allclose(
m.params, N.array([3.20996572, -1.7709542], dtype='float64'))
assert N.allclose(
m.perror, N.array([0.02221018, 0.01893756], dtype='float64'))
chisq = (myfunctlin(m.params, x=x, y=y, err=ey)[1] ** 2).sum()
assert N.allclose(
N.array(
[chisq], dtype='float64'), N.array(
[2.756284983], dtype='float64'))
assert m.dof == 8
return
def test_linfit_bounds():
p0, parinfo, fa = generate_toy_model()
# This is a bad bound, but its to test it works
parinfo[0]["limits"] = [1.5, 1.8]
parinfo[0]["limited"] = [0, 1]
m = mpfit(myfunctlin, p0, parinfo=parinfo, functkw=fa)
res = m.optimize_result
assert res.success
assert res.dof == 8
assert res.status == 1
assert m.params[0] >= 1.5 and m.params[0] <= 1.8
np.testing.assert_allclose(m.params,
np.array([1.8, -1.86916384]),
rtol=5e-7)
np.testing.assert_allclose(m.perror,
np.array([0.0, 0.01887426]),
rtol=5e-7)
chisq = np.sum(
myfunctlin(m.params, x=fa["x"], y=fa["y"], err=fa["err"])[1]**2)
print(chisq)
np.testing.assert_allclose(np.array([chisq]),
np.array([4032.830936495]),
rtol=5e-7)
def test_linfit():
x = N.array([-1.7237128E+00, 1.8712276E+00, -9.6608055E-01,
-2.8394297E-01, 1.3416969E+00, 1.3757038E+00,
-1.3703436E+00, 4.2581975E-02, -1.4970151E-01,
8.2065094E-01])
y = N.array([1.9000429E-01, 6.5807428E+00, 1.4582725E+00,
2.7270851E+00, 5.5969253E+00, 5.6249280E+00,
0.787615, 3.2599759E+00, 2.9771762E+00,
4.5936475E+00])
ey = 0.07 * N.ones(y.shape, dtype='float64')
p0 = N.array([1.0, 1.0], dtype='float64') # initial conditions
pactual = N.array([3.2, 1.78]) # actual values used to make data
parbase = {'value': 0., 'fixed': 0, 'limited': [0, 0], 'limits': [0., 0.]}
parinfo = []
for i in range(len(pactual)):
parinfo.append(copy.deepcopy(parbase))
for i in range(len(pactual)):
parinfo[i]['value'] = p0[i]
fa = {'x': x, 'y': y, 'err': ey}
m = mpfit(myfunctlin, p0, parinfo=parinfo, functkw=fa)
if m.status <= 0:
print('error message = ', m.errmsg)
assert N.allclose(
m.params, N.array([3.20996572, -1.7709542], dtype='float64'))
assert N.allclose(
m.perror, N.array([0.02221018, 0.01893756], dtype='float64'))
chisq = (myfunctlin(m.params, x=x, y=y, err=ey)[1] ** 2).sum()
assert N.allclose(
N.array(
[chisq], dtype='float64'), N.array(
[2.756284983], dtype='float64'))
assert m.dof == 8
return
def test_rosenbrock():
p0 = N.array([-1, 1.], dtype='float64') # initial conditions
pactual = N.array([1., 1.]) # actual minimum of the rosenbrock function
m = mpfit(myfunctrosenbrock, p0)
if m.status <= 0:
print 'error message = ', m.errmsg
assert m.status > 0
assert N.allclose(m.params, pactual)
assert N.allclose(m.fnorm, 0)
return
def test_rosenbrock():
p0 = N.array([-1, 1.], dtype='float64') # initial conditions
pactual = N.array([1., 1.]) # actual minimum of the rosenbrock function
m = mpfit(myfunctrosenbrock, p0)
if m.status <= 0:
print('error message = ', m.errmsg)
assert m.status > 0
assert N.allclose(m.params, pactual)
assert N.allclose(m.fnorm, 0)
return
def test_rosenbrock():
p0 = np.array([-1.0, 1.0])
m = mpfit(myfunctrosenbrock, p0)
res = m.optimize_result
assert isinstance(res, OptimizeResult)
assert res.success
exp_param = np.array([1.0, 1.0])
exp_fnorm = 0.0
exp_error = np.array([0.70710678, 1.41598024])
np.testing.assert_allclose(res.x, exp_param, rtol=5e-7)
np.testing.assert_allclose(res.fnorm, exp_fnorm, rtol=5e-7)
np.testing.assert_allclose(res.perror, exp_error, rtol=5e-7)
def test_linfit():
p0, parinfo, fa = generate_toy_model()
m = mpfit(myfunctlin, p0, parinfo=parinfo, functkw=fa)
res = m.optimize_result
assert res.success
assert res.dof == 8
assert res.status == 1
np.testing.assert_allclose(m.params,
np.array([3.20996572, -1.7709542]),
rtol=5e-7)
np.testing.assert_allclose(m.perror,
np.array([0.02221018, 0.01893756]),
rtol=5e-7)
chisq = np.sum(
myfunctlin(m.params, x=fa["x"], y=fa["y"], err=fa["err"])[1]**2)
np.testing.assert_allclose(np.array([chisq]),
np.array([2.756284983]),
rtol=5e-7)
def test_rosenbrock_error():
# Initial conditions
p0 = np.array([-1.0, 1.0])
m = mpfit(myfunctrosenbrock, p0, maxiter=1)
assert m.status == 5
np.testing.assert_allclose(m.params, p0, rtol=5e-7)
def boundedfit(
model,
old_p1,
bounded=True,
update_plot=False,
return_info=True,
**kwargs,
):
'''
Replica of bounded fit function from hyperspy/model.py, massively
simplified, with multiple assumed variable values.
Arguments:
model -- The model to be fitted, this needs to be passed as fit(),
in this case, is not a member function of the Model class
as it is in model.py
old_p1 -- Inherited parameter set.
bounded -- Flag indicating whether bounded fit function should be used.
Not strictly necessary for this function but a remnant from
the hyperspy implementation where bounded fit is enclosed as
an option inside fit() in hyperspy/model.py.
update_plot -- Flag indicating whether to update any potential interactive
plot
return_info -- Flag indicating whether or not to return results from
function
**kwargs -- Keyword arguments to pass to mpfit() from
hyperspy.external.mpfit.mpfit.
Output:
model.fit_output -- If return_info=True, return the output of the
fit function, mpfit()
'''
# Context manager, not quite sure what this does
cm = (model.suspend_update if (update_plot != model._plot_active)
and not update_plot else dummy_context_manager)
# Bind existing parameters inside their prescribed limits
if bounded:
model.ensure_parameters_in_bounds()
with cm(update_on_resume=True):
model.p_std = None
model.p0 = old_p1 # Set existing p0 to inherited p0
old_p0 = model.p0 # For checking fit function success later
weights = model._convert_variance_to_weights()
args = (model.signal()[np.where(model.channel_switches)], weights)
model._set_mpfit_parameters_info(bounded=bounded)
auto_deriv = 1
# Actual optimization function using hyperspy.external.mpfit.mpfit
# The model._errfunc4mpfit calls a function which is defined in
# ScalableReferencePattern.function() in pyxem/components/scalable_reference_pattern.py
# model.mpfit_parinfo contains information about the limits set for the parameters
res = mpfit(
model._errfunc4mpfit,
model.p0[:],
parinfo=model.mpfit_parinfo,
functkw={
"y": model.signal()[model.channel_switches],
"weights": weights,
},
autoderivative=auto_deriv,
quiet=1,
**kwargs,
)
# Create an object, OptimizeResult, to store optimization results
model.fit_output = OptimizeResult(
x=res.params,
covar=res.covar,
perror=res.perror,
nit=res.niter,
nfev=res.nfev,
success=(res.status > 0) and (res.status != 5),
status=res.status,
message=res.errmsg,
debug=res.debug,
dof=res.dof,
fnorm=res.fnorm,
)
model.p0 = model.fit_output.x # Update parameter set in model
ysize = len(model.fit_output.x) + model.fit_output.dof
cost = model.fit_output.fnorm
pcov = model.fit_output.perror**2 # Covariance of parameter fit results
# Standard deviation of parameter fit results
model.p_std = model._calculate_parameter_std(pcov, cost, ysize)
# If only 1 parameter is in set p0 then turn into a list of
# parameters 1 element long (for the sake of compatibility)
if np.iterable(model.p0) == 0:
model.p0 = (model.p0, )
# Store parameter values in a map corresponding to the image data used
# to create the model
model._fetch_values_from_p0(p_std=model.p_std)
model.store_current_values()
model._calculate_chisq()
model._set_current_degrees_of_freedom()
# Check that the parameters have been changed by the fitting
if np.any(old_p0 != model.p0):
model.events.fitted.trigger(model)
# Success check and error message output
success = model.fit_output.get("success", None)
if success is False:
message = model.fit_output.get("message", "Unknown reason")
_logger.warning(
f"`m.fit()` did not exit successfully. Reason: {message}")
if return_info:
return model.fit_output
else:
return None
