def create_loss():
bounds = (0.1, 3.0)
obs = zfit.Space("x", limits=bounds)
# Data and signal
np.random.seed(0)
tau = -2.0
beta = -1 / tau
bkg = np.random.exponential(beta, 300)
peak = np.random.normal(1.2, 0.1, 80)
data = np.concatenate((bkg, peak))
data = data[(data > bounds[0]) & (data < bounds[1])]
N = len(data)
data = zfit.data.Data.from_numpy(obs=obs, array=data)
mean = zfit.Parameter("mean", 1.2, 0.5, 2.0)
sigma = zfit.Parameter("sigma", 0.1, 0.02, 0.2)
lambda_ = zfit.Parameter("lambda", -2.0, -4.0, -1.0)
Nsig = zfit.Parameter("Ns", 20.0, -20.0, N)
Nbkg = zfit.Parameter("Nbkg", N, 0.0, N * 1.1)
signal = zfit.pdf.Gauss(obs=obs, mu=mean,
sigma=sigma).create_extended(Nsig)
background = zfit.pdf.Exponential(obs=obs,
lambda_=lambda_).create_extended(Nbkg)
tot_model = zfit.pdf.SumPDF([signal, background])
loss = ExtendedUnbinnedNLL(model=tot_model, data=data)
return loss, mean
def create_loss():
with tf.variable_scope("func1"):
a_param = zfit.Parameter("variable_a15151",
1.5,
-1.,
20.,
step_size=ztf.constant(0.1))
b_param = zfit.Parameter("variable_b15151", 3.5)
c_param = zfit.Parameter("variable_c15151", -0.04)
obs1 = zfit.Space(obs='obs1', limits=(-2.4, 9.1))
# load params for sampling
a_param.load(true_a)
b_param.load(true_b)
c_param.load(true_c)
gauss1 = zfit.pdf.Gauss(mu=a_param, sigma=b_param, obs=obs1)
exp1 = zfit.pdf.Exponential(lambda_=c_param, obs=obs1)
sum_pdf1 = 0.9 * gauss1 + exp1
sampled_data = sum_pdf1.create_sampler(n=15000)
sampled_data.resample()
loss = zfit.loss.UnbinnedNLL(model=sum_pdf1,
data=sampled_data,
fit_range=obs1)
return loss, (a_param, b_param, c_param)
def test_set_values():
upper1 = 5.33
lower1 = 0.
param1 = zfit.Parameter('param1', 2., lower1, upper1)
param1.set_value(lower1)
assert pytest.approx(zfit.run(param1.value()), lower1)
param1.set_value(upper1)
assert pytest.approx(zfit.run(param1.value()), upper1)
with pytest.raises(ValueError):
param1.set_value(lower1 - 0.001)
with pytest.raises(ValueError):
param1.set_value(upper1 + 0.001)
fitresult = create_fitresult((zfit.minimize.Minuit, {}, True))
result = fitresult['result']
param_a = fitresult['a_param']
param_b = fitresult['b_param']
param_c = fitresult['c_param']
val_a = fitresult['a']
val_b = fitresult['b']
val_c = fitresult['c']
with pytest.raises(ValueError):
param_b.set_value(999)
param_c.assign(9999)
zfit.param.set_values([param_c, param_b], values=result)
assert param_a.value() == val_a
assert param_b.value() == val_b
assert param_c.value() == val_c
param_d = zfit.Parameter("param_d", 12)
with pytest.raises(ValueError):
zfit.param.set_values([param_d], result)
def test_with_gauss_qtilde(n, min_x):
sigma_x = 0.032
minimizer = Minuit()
bounds = (-10, 10)
obs = zfit.Space("x", limits=bounds)
mean = zfit.Parameter("mean", n * sigma_x)
sigma = zfit.Parameter("sigma", 1.0)
model = zfit.pdf.Gauss(obs=obs, mu=mean, sigma=sigma)
data = model.sample(n=1000)
nll = UnbinnedNLL(model=model, data=data)
minimum = minimizer.minimize(loss=nll)
minimum.hesse()
x = minimum.params[mean]["value"]
x_err = minimum.params[mean]["minuit_hesse"]["error"]
x_min = x - x_err * 3
x_max = x + x_err * 3
x_min = max([x_min, min_x])
poinull = POIarray(mean, np.linspace(x_min, x_max, 50))
calculator = AsymptoticCalculator(nll, minimizer)
ci = ConfidenceInterval(calculator, poinull, qtilde=True)
ci.interval(alpha=0.05, printlevel=1)
def create_loss():
bounds = (0.1, 3.0)
obs = zfit.Space('x', limits=bounds)
# Data and signal
np.random.seed(0)
tau = -2.0
beta = -1 / tau
bkg = np.random.exponential(beta, 300)
peak = np.random.normal(1.2, 0.1, 25)
data = np.concatenate((bkg, peak))
data = data[(data > bounds[0]) & (data < bounds[1])]
N = len(data)
data = zfit.data.Data.from_numpy(obs=obs, array=data)
lambda_ = zfit.Parameter("lambda", -2.0, -4.0, -1.0)
Nsig = zfit.Parameter("Ns", 20., -20., N)
Nbkg = zfit.Parameter("Nbkg", N, 0., N * 1.1)
signal = Nsig * zfit.pdf.Gauss(obs=obs, mu=1.2, sigma=0.1)
background = Nbkg * zfit.pdf.Exponential(obs=obs, lambda_=lambda_)
tot_model = signal + background
loss = ExtendedUnbinnedNLL(model=[tot_model], data=[data], fit_range=[obs])
return loss, (Nsig, Nbkg)
def create_loss(n=15000, weights=None):
avalue = 1.5
a_param = zfit.Parameter("variable_a15151", avalue, -1., 20.,
step_size=z.constant(0.1))
a_param.init_val = avalue
bvalue = 3.5
b_param = zfit.Parameter("variable_b15151", bvalue, 0, 20)
b_param.init_val = bvalue
cvalue = -0.04
c_param = zfit.Parameter("variable_c15151", cvalue, -1, 0.)
c_param.init_val = cvalue
obs1 = zfit.Space(obs='obs1', limits=(-2.4, 9.1))
# load params for sampling
a_param.set_value(true_a)
b_param.set_value(true_b)
c_param.set_value(true_c)
gauss1 = zfit.pdf.Gauss(mu=a_param, sigma=b_param, obs=obs1)
exp1 = zfit.pdf.Exponential(lam=c_param, obs=obs1)
sum_pdf1 = zfit.pdf.SumPDF((gauss1, exp1), 0.7)
sampled_data = sum_pdf1.create_sampler(n=n)
sampled_data.resample()
if weights is not None:
sampled_data.set_weights(weights)
loss = zfit.loss.UnbinnedNLL(model=sum_pdf1, data=sampled_data)
return loss, (a_param, b_param, c_param)
def test_with_gauss_fluctuations():
x_true = -2.0
minimizer = Minuit()
bounds = (-10, 10)
obs = zfit.Space("x", limits=bounds)
mean = zfit.Parameter("mean", 0)
sigma = zfit.Parameter("sigma", 1.0)
model = zfit.pdf.Gauss(obs=obs, mu=mean, sigma=sigma)
npzfile = f"{notebooks_dir}/toys/FC_toys_{x_true}.npz"
data = zfit.data.Data.from_numpy(obs=obs, array=np.load(npzfile)["x"])
nll = UnbinnedNLL(model=model, data=data)
minimum = minimizer.minimize(loss=nll)
minimum.hesse()
toys_fname = f"{notebooks_dir}/toys/FC_toys_{x_true}.yml"
calculator = FrequentistCalculator.from_yaml(toys_fname, minimum,
minimizer)
keys = np.unique([k[0].value for k in calculator.keys()])
keys.sort()
poinull = POIarray(mean, keys)
ci = ConfidenceInterval(calculator, poinull, qtilde=False)
with pytest.warns(UserWarning):
ci.interval(alpha=0.05, printlevel=0)
ci = ConfidenceInterval(calculator, poinull, qtilde=True)
ci.interval(alpha=0.05, printlevel=0)
def test_cb_integral():
obs = zfit.Space("x", limits=bounds)
mu_ = zfit.Parameter("mu_cb5", mu)
sigma_ = zfit.Parameter("sigma_cb5", sigma)
alphal_ = zfit.Parameter("alphal_cb5", alphal)
nl_ = zfit.Parameter("nl_cb5", nl)
cbl = CrystalBall(obs=obs, mu=mu_, sigma=sigma_, alpha=alphal_, n=nl_)
int_limits = (-1, 3)
integral_numeric = cbl.numeric_integrate(limits=int_limits, norm=False)
integral = cbl.analytic_integrate(limits=int_limits, norm=False)
integral_numeric = zfit.run(integral_numeric)
integral = zfit.run(integral)
assert pytest.approx(integral_numeric, integral, 1e-5)
rnd_limits = sorted(np.random.uniform(*bounds, 13))
integrals = []
for low, up in zip(rnd_limits[:-1], rnd_limits[1:]):
integrals.append(cbl.integrate((low, up), norm=False))
integral = np.sum(integrals)
integral_full = zfit.run(cbl.integrate(bounds, norm=False))
assert pytest.approx(integral_full, integral)
def test_conv_simple(interpolation):
n_points = 2432
obs = zfit.Space("obs1", limits=(-5, 5))
param1 = zfit.Parameter('param1', -3)
param2 = zfit.Parameter('param2', 0.3)
gauss = zfit.pdf.Gauss(0., param2, obs=obs)
func1 = zfit.pdf.Uniform(param1, param2, obs=obs)
func2 = zfit.pdf.Uniform(-1.2, -1, obs=obs)
func = zfit.pdf.SumPDF([func1, func2], 0.5)
conv = zfit.pdf.FFTConvPDFV1(func=func, kernel=gauss, n=100, interpolation=interpolation)
if interpolation == 'spline:5':
assert conv._conv_spline_order == 5
elif interpolation == 'spline:3':
assert conv._conv_spline_order == 3
x = tf.linspace(-5., 5., n_points)
probs = conv.pdf(x=x)
# true convolution
true_conv = true_conv_np(func, gauss, obs, x, xkernel=tf.linspace(*obs.limit1d, num=n_points))
integral = conv.integrate(limits=obs)
probs_np = probs.numpy()
np.testing.assert_allclose(probs, true_conv, rtol=0.01, atol=0.01)
assert pytest.approx(1, rel=1e-3) == integral.numpy()
assert len(probs_np) == n_points
def plot_conv_comparison():
# test special properties here
n_point_plotting = 2000
obs = zfit.Space("obs1", limits=(-5, 5))
param1 = zfit.Parameter("param1", -3)
param2 = zfit.Parameter("param2", 0.3)
gauss1 = zfit.pdf.Gauss(0.0, param2, obs=obs)
func1 = zfit.pdf.Uniform(param1, param2, obs=obs)
func2 = zfit.pdf.Uniform(-1.2, -1, obs=obs)
func = zfit.pdf.SumPDF([func1, func2], 0.5)
n_points_conv = 50
conv_lin = zfit.pdf.FFTConvPDFV1(func=func,
kernel=gauss1,
n=n_points_conv,
interpolation="linear")
conv_spline = zfit.pdf.FFTConvPDFV1(func=func,
kernel=gauss1,
n=n_points_conv,
interpolation="spline")
x = tf.linspace(-5.0, 3.0, n_point_plotting)
probs_lin = conv_lin.pdf(x=x)
probs_spline = conv_spline.pdf(x=x)
plt.figure()
plt.plot(x, probs_lin, label="linear")
plt.plot(x, probs_spline, label="spline")
plt.legend()
plt.title(f"FFT Conv with interpolation: {n_points_conv} points")
plt.show(block=True)
def test_gaussian_constraint_orderbug2(
): # as raised in #162, failed before fixing
param1 = zfit.Parameter("param1", 1500)
param5 = zfit.Parameter("param2", 0.5)
param2 = zfit.Parameter("param3", 1.0)
param3 = zfit.Parameter("param4", 1.0)
param4 = zfit.Parameter("param5", 1.0)
constraint = {
"params": [param1, param2, param3, param4, param5],
"mu": [1500, 1.0, 1.0, 1.0, 0.5],
"sigma": [0.05 * 1500, 0.001, 0.01, 0.1, 0.05 * 0.5]
}
constr1 = zfit.constraint.GaussianConstraint(**constraint)
# param_vals = [1500, 1.0, 1.0, 1.0, 0.5]
constraint['params'] = zfit.run(constraint['params'])
true_val = true_gauss_constr_value(**constraint)
value_tensor = constr1.value()
constr_np = zfit.run(value_tensor)
assert constr_np == pytest.approx(true_val)
assert true_val < 1000
assert true_val == pytest.approx(
-8.592,
abs=0.1) # if failing, change value. Hardcoded for additional layer
def create_loss(n=15000):
a_param = zfit.Parameter("variable_a15151",
1.5,
-1.,
20.,
step_size=z.constant(0.1))
b_param = zfit.Parameter("variable_b15151", 3.5, 0, 20)
c_param = zfit.Parameter("variable_c15151", -0.04, -1, 0.)
obs1 = zfit.Space(obs='obs1', limits=(-2.4, 9.1))
# load params for sampling
a_param.set_value(true_a)
b_param.set_value(true_b)
c_param.set_value(true_c)
gauss1 = zfit.pdf.Gauss(mu=a_param, sigma=b_param, obs=obs1)
exp1 = zfit.pdf.Exponential(lambda_=c_param, obs=obs1)
sum_pdf1 = zfit.pdf.SumPDF((gauss1, exp1), 0.7)
sampled_data = sum_pdf1.create_sampler(n=n)
sampled_data.resample()
loss = zfit.loss.UnbinnedNLL(model=sum_pdf1, data=sampled_data)
return loss, (a_param, b_param, c_param)
def test_gaussian_constraint_orderbug2(
): # as raised in #162, failed before fixing
param1 = zfit.Parameter("param1", 1500)
param5 = zfit.Parameter("param2", 0.5)
param2 = zfit.Parameter("param3", 1.0)
param3 = zfit.Parameter("param4", 1.0)
param4 = zfit.Parameter("param5", 1.0)
constraint = {
"params": [param1, param2, param3, param4, param5],
"observation": [1500, 1.0, 1.0, 1.0, 0.5],
"uncertainty": [0.05 * 1500, 0.001, 0.01, 0.1, 0.05 * 0.5],
}
constr1 = GaussianConstraint(**constraint)
# param_vals = [1500, 1.0, 1.0, 1.0, 0.5]
constraint["x"] = [m.numpy() for m in constraint["params"]]
true_val = true_gauss_constr_value(x=constraint["x"],
mu=constraint["observation"],
sigma=constraint["uncertainty"])
value_tensor = constr1.value()
constr_np = value_tensor.numpy()
assert constr_np == pytest.approx(true_val)
assert true_val < 1000
assert true_val == pytest.approx(
-8.592,
abs=0.1) # if failing, change value. Hardcoded for additional layer
def create_test_gauss1():
mu2 = zfit.Parameter("mu2_sampling1", mu_true, mu_true - 2.0, mu_true + 7.0)
sigma2 = zfit.Parameter(
"sigma2_sampling1", sigma_true, sigma_true - 10.0, sigma_true + 5.0
)
test_gauss1 = TmpGaussian(name="test_gauss1", mu=mu2, sigma=sigma2, obs=obs1)
return test_gauss1, mu2, sigma2
def create_mu_sigma_2(nameadd=""):
import zfit
mu2 = zfit.Parameter("mu2" + nameadd, mu_true, mu_true - 2.0, mu_true + 7.0)
sigma2 = zfit.Parameter(
"sigma2" + nameadd, sigma_true, sigma_true - 10.0, sigma_true + 5.0
)
return mu2, sigma2
def create_loss_1():
obs = zfit.Space("x", limits=(0.1, 2.0))
data = zfit.data.Data.from_numpy(obs=obs, array=np.random.normal(1.2, 0.1, 10000))
mean = zfit.Parameter("mu", 1.2)
sigma = zfit.Parameter("sigma", 0.1)
model = zfit.pdf.Gauss(obs=obs, mu=mean, sigma=sigma)
loss = UnbinnedNLL(model=model, data=data)
return loss
def create_gauss3(nameadd=""):
import zfit
obs1 = zfit.Space('obs1', (low, high))
mu3 = zfit.Parameter("mu3" + nameadd, mu_true, mu_true - 2., mu_true + 7.)
sigma3 = zfit.Parameter("sigma3" + nameadd, sigma_true, sigma_true - 10.,
sigma_true + 5.)
gauss3 = zfit.pdf.Gauss(mu=mu3, sigma=sigma3, obs=obs1)
return gauss3
def create_initial_model(obs, sample):
# Crystal Ball
mu = zfit.Parameter(f"mu_{sample}", 80., 60., 120.)
sigma = zfit.Parameter(f'sigma_{sample}', 8., 1., 100.)
alpha = zfit.Parameter(f'alpha_{sample}', -.5, -10., 0.)
n = zfit.Parameter(f'n_{sample}', 120., 0.01, 200.)
model = zfit.pdf.CrystalBall(obs=obs, mu=mu, sigma=sigma, alpha=alpha, n=n)
return model
def test_overloaded_operators():
param1 = zfit.Parameter('param1', 5)
param2 = zfit.Parameter('param2', 3)
param_a = ComposedParameter('param_a', lambda p1: p1 * 4, params=param1)
param_b = ComposedParameter('param_b', lambda p2: p2, params=param2)
param_c = param_a * param_b
assert not isinstance(param_c, zfit.Parameter)
param_d = ComposedParameter("param_d", lambda pa, pb: pa + pa * pb ** 2, params=[param_a, param_b])
param_d_val = param_d.numpy()
assert param_d_val == (param_a + param_a * param_b ** 2).numpy()
def create_test_pdf_overriden_gauss1():
mu2 = zfit.Parameter("mu2_sampling1", mu_true, mu_true - 2., mu_true + 7.)
sigma2 = zfit.Parameter("sigma2_sampling1", sigma_true, sigma_true - 10.,
sigma_true + 5.)
test_gauss1 = TmpGaussianPDFNonNormed(name="test_pdf_nonnormed_gauss1",
mu=mu2,
sigma=sigma2,
obs=obs1)
return test_gauss1, mu2, sigma2
def create_loss():
obs = zfit.Space('x', limits=(0.1, 2.0))
data = zfit.data.Data.from_numpy(obs=obs, array=np.random.normal(1.2, 0.1, 10000))
mean = zfit.Parameter("mu", true_mu)
sigma = zfit.Parameter("sigma", true_sigma)
model = zfit.pdf.Gauss(obs=obs, mu=mean, sigma=sigma)
loss = UnbinnedNLL(model=[model], data=[data], fit_range=[obs])
return loss, (mean, sigma)
def create_gauss1():
mu = zfit.Parameter("mu_sampling1", mu_true, mu_true - 2., mu_true + 7.)
sigma = zfit.Parameter("sigma_sampling1", sigma_true, sigma_true - 10.,
sigma_true + 5.)
gauss_params1 = zfit.pdf.Gauss(mu=mu,
sigma=sigma,
obs=obs1,
name="gauss_params1_sampling1")
return gauss_params1, mu, sigma
def test_numerical_hessian():
param1 = zfit.Parameter('param1', 4.)
param2 = zfit.Parameter('param2', 5.)
param3 = zfit.Parameter('param3', 2.)
def func1():
return param1 * param2 ** 2 + param3 ** param1
num_hessian = numerical_hessian(func1, params=[param1, param2, param3])
tf_hessian = autodiff_hessian(func1, params=[param1, param2, param3])
np.testing.assert_allclose(num_hessian, tf_hessian, rtol=1e-5, atol=1e-10)
def create_params2(nameadd=""):
mu2 = zfit.Parameter("mu25" + nameadd,
z.to_real(mu_true) - 0.2, mu_true - 1.0,
mu_true + 1.0)
sigma2 = zfit.Parameter(
"sigma25" + nameadd,
z.to_real(sigma_true) - 0.3,
sigma_true - 2.0,
sigma_true + 2.0,
)
return mu2, sigma2
def test_numerical_gradient():
param1 = zfit.Parameter('param1', 4.)
param2 = zfit.Parameter('param2', 5.)
param3 = zfit.Parameter('param3', 2.)
def func1():
return param1 * param2**2 + param3**param1
num_gradients = numerical_gradient(func1, params=[param1, param2, param3])
tf_gradients = autodiff_gradient(func1, params=[param1, param2, param3])
np.testing.assert_allclose(num_gradients, tf_gradients)
def create_params1(nameadd=""):
mu1 = zfit.Parameter("mu1" + nameadd,
z.to_real(mu_true) - 0.2, mu_true - 5.0,
mu_true + 5.0)
sigma1 = zfit.Parameter(
"sigma1" + nameadd,
z.to_real(sigma_true) - 0.3,
sigma_true - 4.0,
sigma_true + 4.0,
)
return mu1, sigma1
def create_params3(nameadd=""):
mu3 = zfit.Parameter("mu35" + nameadd,
z.to_real(mu_true) - 0.2, mu_true - 1.0,
mu_true + 1.0)
sigma3 = zfit.Parameter(
"sigma35" + nameadd,
z.to_real(sigma_true) - 0.3,
sigma_true - 2.0,
sigma_true + 2.0,
)
yield3 = zfit.Parameter("yield35" + nameadd, yield_true + 300, 0, 10000000)
return mu3, sigma3, yield3
def test_gaussian_constraint_matrix():
param1 = zfit.Parameter("Param1", 5)
param2 = zfit.Parameter("Param2", 6)
params = [param1, param2]
mu = [3., 6.1]
sigma = np.array([[1, 0.3], [0.3, 0.5]])
constr = GaussianConstraint(params=params, mu=mu, sigma=sigma)
constr_np = zfit.run(constr.value())
assert constr_np == pytest.approx(3.989638)
assert constr.get_dependents() == set(params)
def create_gauss1(nameadd=""):
import zfit
obs1 = zfit.Space("obs1", (low, high))
mu = zfit.Parameter("mu" + nameadd, mu_true, mu_true - 2.0, mu_true + 7.0)
sigma = zfit.Parameter(
"sigma" + nameadd, sigma_true, sigma_true - 10.0, sigma_true + 5.0
)
gauss_params1 = zfit.pdf.Gauss(mu=mu, sigma=sigma, obs=obs1, name="gauss_params1")
return gauss_params1
def test_gaussian_constraint_shape_errors():
param1 = zfit.Parameter("Param1", 5)
param2 = zfit.Parameter("Param2", 6)
with pytest.raises(ShapeIncompatibleError):
GaussianConstraint([param1, param2], mu=[4, 2, 3], sigma=5)
with pytest.raises(ShapeIncompatibleError):
GaussianConstraint([param1, param2], mu=[4, 2], sigma=5)
with pytest.raises(ShapeIncompatibleError):
GaussianConstraint([param1, param2], mu=2, sigma=[1, 4])
with pytest.raises(ShapeIncompatibleError):
GaussianConstraint(param1, mu=[4, 2], sigma=[2, 3])
