def test_type():
"Check that the function returns a valid model type"
model = dl.dd_gauss
newmodel = relate(model, width = lambda mean: mean/2)
assert isinstance(newmodel,Model)
def test_relate_signature():
"Check the model signature after relate"
model = dl.dd_gauss2
newmodel = relate(model,mean1= lambda mean2: mean2)
assert newmodel.signature==['r', 'mean1', 'width1', 'width2', 'amp1', 'amp2']
def test_Nparam_lin():
"Check that the related model has the right number of parameters"
model = dl.dd_gauss2
newmodel = relate(model, width1 = lambda mean1: mean1/2)
assert newmodel.Nlin == model.Nlin
def test_relate_one_nonlinear():
"Check that the method works correctly for one related parameter"
model = dl.dd_gauss
x = np.linspace(0,10,400)
ref = model(x,4,0.2)
newmodel = relate(model, width = lambda mean: mean/20)
response = newmodel(x,4)
assert np.allclose(response,ref)
def test_relate_conflict_order():
"Check that even if the user specifies a conflicted order, it runs"
model = dl.dd_gauss2
x = np.linspace(0,10,400)
ref = model(x,4,0.2,5,0.4,1,1)
# No conflict, there parameters are ordered correctly
newmodel = relate(model,
width2=lambda width1: width1*2,
width1=lambda mean1: mean1/20,)
# Conflict: width1 is deleted first, but width2 depends on it
conflictmodel = relate(model,
width1=lambda mean1: mean1/20,
width2=lambda width1: width1*2)
response1 = newmodel(r=x,mean1=4,mean2=5,amp1=1,amp2=1)
response2 = conflictmodel(r=x,mean1=4,mean2=5,amp1=1,amp2=1)
assert np.allclose(response1,ref) and np.allclose(response2,ref)
def test_relate_fit():
"Check that the method works correctly for one related parameter"
model = dl.dd_gauss
x = np.linspace(0,10,400)
ref = model(x,4,0.2)
newmodel = relate(model, width = lambda mean: mean/20)
result = fit(newmodel,ref,x)
assert np.allclose(result.model,ref)
def test_relate_two_nonlinear():
"Check that the method works correctly for two related parameters"
model = dl.dd_gauss2
x = np.linspace(0,10,400)
ref = model(x,3,0.2,6,0.1,1,1)
newmodel = relate(model,
mean1 = lambda mean2: mean2/2,
width1 = lambda width2: width2*2)
response = newmodel(r=x,mean2=6,width2=0.1,amp1=1,amp2=1)
assert np.allclose(response,ref)
def test_relate_addednonlinear():
"Check that a parameter can be related with an added nonlinear parameter"
model = deepcopy(dl.dd_gauss)
model.addlinear('amp')
x = np.linspace(0,10,400)
ref = model(x,4,0.2,5)
model.addnonlinear('factor')
newmodel = relate(model, width = lambda factor: factor/5)
response = newmodel(r=x,mean=4,factor=1,amp=5)
assert np.allclose(response,ref)
def test_param_names():
"Check that the related model has the adjusted parameter names"
model = dl.dd_gauss
newmodel = relate(model, width = lambda mean: mean/2)
assert all([ str in newmodel._parameter_list() for str in ['mean'] ])
def test_Nparam_list():
"Check that the related model has the right number of parameters"
model = dl.dd_gauss3
newmodel = relate(model, width1 = lambda mean1: mean1/2)
assert newmodel.Nparam == len(newmodel._parameter_list())
def test_Nparam_nonlin():
"Check that the output model has the right number of parameters"
model = dl.dd_gauss
newmodel = relate(model, width = lambda mean: mean/2)
assert newmodel.Nnonlin == model.Nnonlin - 1
def test_preserve_original():
"Check that the original input model is not changed by the function"
model = dl.dd_gauss
_ = relate(model, width = lambda mean: mean/2)
assert model._parameter_list() == ['mean','width']