def test_opts(pars):
ds = MyDataset(pars)
factors, info, minuit = optimize_iminuit(
function=ds.fcn, parameters=pars, migrad_opts={"ncall": 20}, tol=1.0, strategy=2
)
assert info["nfev"] == 29
assert minuit.tol == 1.0
assert minuit.strategy == 2
def test_migrad_opts(pars):
ds = MyDataset(pars)
kwargs = {}
kwargs["migrad_opts"] = {"ncall": 20}
factors, info, minuit = optimize_iminuit(function=ds.fcn,
parameters=pars,
**kwargs)
assert info["nfev"] == 20
def test_iminuit_stepsize():
ds = MyDataset()
pars = ds.models.parameters
factors, info, minuit = optimize_iminuit(function=ds.fcn, parameters=pars)
assert info["success"]
assert_allclose(ds.fcn(), 0, atol=1e-5)
assert_allclose(pars["x"].value, 2, rtol=1e-3)
assert_allclose(pars["y"].value, 3e5, rtol=1e-3)
assert_allclose(pars["z"].value, 4e-5, rtol=2e-2)
def test_iminuit_frozen(pars):
ds = MyDataset(pars)
pars["y"].frozen = True
factors, info, minuit = optimize_iminuit(function=ds.fcn, parameters=pars)
assert info["success"]
assert_allclose(pars["x"].value, 2, rtol=1e-4)
assert_allclose(pars["y"].value, 3.1e5)
assert_allclose(pars["z"].value, 4.0e-5, rtol=1e-4)
assert_allclose(ds.fcn(), 0.111112, rtol=1e-5)
def test_iminuit_stepsize(pars):
ds = MyDataset(pars)
pars.covariance = np.diag([0.2, 3e4, 4e-6]) ** 2
factors, info, minuit = optimize_iminuit(function=ds.fcn, parameters=pars)
assert info["success"]
assert_allclose(ds.fcn(), 0, atol=1e-5)
assert_allclose(pars["x"].value, 2, rtol=1e-3)
assert_allclose(pars["y"].value, 3e5, rtol=1e-3)
assert_allclose(pars["z"].value, 4e-5, rtol=2e-2)
def test_iminuit_confidence(pars):
ds = MyDataset(pars)
factors, info, minuit = optimize_iminuit(function=ds.fcn, parameters=pars)
assert_allclose(ds.fcn(), 0, atol=1e-5)
par = pars["x"]
par.min, par.max = 0, 10
result = confidence_iminuit(minuit=minuit, parameters=pars, parameter=par, sigma=1)
assert result["success"]
assert_allclose(result["errp"], 0.2)
assert_allclose(result["errn"], 0.2)
def test_iminuit_basic(pars):
ds = MyDataset(pars)
factors, info, minuit = optimize_iminuit(function=ds.fcn, parameters=pars)
assert info["success"]
assert_allclose(ds.fcn(), 0, atol=1e-5)
# Check the result in parameters is OK
assert_allclose(pars["x"].value, 2, rtol=1e-3)
assert_allclose(pars["y"].value, 3e5, rtol=1e-3)
# Precision of estimate on "z" is very poor (0.040488). Why is it so bad?
assert_allclose(pars["z"].value, 4e-5, rtol=2e-2)
# Check that minuit sees the parameter factors correctly
assert_allclose(factors, [2, 3, 4], rtol=1e-3)
assert_allclose(minuit.values["par_000_x"], 2, rtol=1e-3)
assert_allclose(minuit.values["par_001_y"], 3, rtol=1e-3)
assert_allclose(minuit.values["par_002_z"], 4, rtol=1e-3)
def test_iminuit_limits(pars):
ds = MyDataset(pars)
pars["y"].min = 301000
factors, info, minuit = optimize_iminuit(function=ds.fcn, parameters=pars)
assert info["success"]
# Check the result in parameters is OK
assert_allclose(pars["x"].value, 2, rtol=1e-2)
assert_allclose(pars["y"].value, 301000, rtol=1e-3)
# Check that minuit sees the limit factors correctly
states = minuit.get_param_states()
assert not states[0]["has_limits"]
y = states[1]
assert y["has_limits"]
assert_allclose(y["lower_limit"], 3.01)
def test_iminuit_limits():
ds = MyDataset()
pars = ds.models.parameters
pars["y"].min = 301000
factors, info, minuit = optimize_iminuit(function=ds.fcn, parameters=pars)
assert info["success"]
# Check the result in parameters is OK
assert_allclose(pars["x"].value, 2, rtol=1e-2)
assert_allclose(pars["y"].value, 301000, rtol=1e-3)
# Check that minuit sees the limit factors correctly
params = minuit.init_params
assert not params["x"].has_limits
assert params["par_001_y"].has_limits
assert_allclose(params["par_001_y"].lower_limit, 3.01)
assert params["par_001_y"].upper_limit is None
