def test_single_component():
# make the sampling deterministic
_ = np.random.seed(42)
source_size = UniformLikelihood.from_values("ss", 0.0, 400.0)
ncol = UniformLikelihood.from_values("ncol", 0.0, 1e16)
vlsr = UniformLikelihood.from_values("vlsr", 0.0, 10.0)
Tex = GaussianLikelihood.from_values("tex", 5.8, 0.5, 0.0, 10.0)
dV = GaussianLikelihood.from_values("dV", 0.1, 1e-1, 0.0, 0.3)
model = SingleComponent(
source_size,
vlsr,
ncol,
Tex,
dV,
None, # no molecule required
None, # no observation required
)
# check the intialization routine is still working
initial = model.sample_prior()
static = [
1.49816048e02,
9.50714306e00,
7.31993942e15,
5.24405994e00,
1.31890218e-01,
]
assert np.allclose(initial, static)
def test_multi_component():
source_sizes = [
UniformLikelihood.from_values("ss", 0., 400.) for _ in range(4)
]
vlsrs = [UniformLikelihood.from_values("vlsr", 0., 10.) for _ in range(4)]
ncols = [UniformLikelihood.from_values("ncol", 0., 1e16) for _ in range(4)]
Tex = GaussianLikelihood.from_values("tex", 5.8, 0.5, 0., 10.)
dV = GaussianLikelihood.from_values("dV", 0.1, 1e-1, 0., 0.3)
model = MultiComponent(
source_sizes,
vlsrs,
ncols,
Tex,
dV,
None,
)
# should be 14 parameters total; 3 * 4 components + 2
assert len(model) == 14
model._get_components()
# make sure the combined likelihood looks reasonable
parameters = [
100., 96, 20, 45, 5., 5.6, 6.44, 4.3, 1e10, 1e11, 1e9, 1e11, 5.87,
0.09216
]
likelihood = model.compute_prior_likelihood(parameters)
assert np.round(abs(likelihood - 4.579927708578582)) == 0.
def test_likelihoods():
source_size = UniformLikelihood.from_values("ss", 0.0, 400.0)
_ = UniformLikelihood.from_values("ncol", 0.0, 1e16)
_ = UniformLikelihood.from_values("vlsr", 0.0, 10.0)
Tex = GaussianLikelihood.from_values("tex", 5.8, 0.5, 0.0, 10.0)
dV = GaussianLikelihood.from_values("dV", 0.1, 1e-1, 0.0, 0.3)
uniform_test = source_size.ln_likelihood(100.0)
assert uniform_test == 0.0
normal_test = Tex.ln_likelihood(5.87)
assert np.round(np.abs(-0.23559135 - normal_test), 4) == 0.0
fail_test = dV.ln_likelihood(-5.0)
assert not np.isfinite(fail_test)
def test_single_component():
source_size = UniformLikelihood.from_values("ss", 0., 400.)
ncol = UniformLikelihood.from_values("ncol", 0., 1e16)
vlsr = UniformLikelihood.from_values("vlsr", 0., 10.)
Tex = GaussianLikelihood.from_values("tex", 5.8, 0.5, 0., 10.)
dV = GaussianLikelihood.from_values("dV", 0.1, 1e-1, 0., 0.3)
model = SingleComponent(
source_size,
vlsr,
ncol,
Tex,
dV,
None,
)
initial = model.initialize_values()
assert initial == [200., 5., 5e15, 5.8, 0.1]
def test_multi_component():
# make the sampling deterministic
_ = np.random.seed(42)
source_sizes = [
UniformLikelihood.from_values("ss", 0.0, 400.0) for _ in range(4)
]
vlsrs = [
UniformLikelihood.from_values("vlsr", 0.0, 10.0) for _ in range(4)
]
ncols = [
UniformLikelihood.from_values("ncol", 0.0, 1e16) for _ in range(4)
]
Tex = GaussianLikelihood.from_values("tex", 5.8, 0.5, 0.0, 10.0)
dV = GaussianLikelihood.from_values("dV", 0.1, 1e-1, 0.0, 0.3)
model = MultiComponent(
source_sizes,
vlsrs,
ncols,
Tex,
dV,
None,
None,
)
# should be 14 parameters total; 3 * 4 components + 2
assert len(model) == 14
model._get_components()
# check the intialization routine is still working deterministically when asked
initial = model.sample_prior()
static = [
149.81604753894499,
62.39780813448106,
8.233797718320979,
73.36180394137352,
9.50714306409916,
0.5808361216819946,
9.699098521619943,
3.0424224295953772,
7319939418114051.0,
8661761457749352.0,
8324426408004217.0,
5247564316322378.0,
4.937541083743484,
0.043771247075902735,
]
assert np.allclose(initial, static)
# make sure the combined likelihood looks reasonable
parameters = [
100.0,
96,
20,
45,
5.0,
5.6,
6.44,
4.3,
1e10,
1e11,
1e9,
1e11,
5.87,
0.09216,
]
likelihood = model.compute_prior_likelihood(parameters)
assert np.round(abs(likelihood - 4.579927708578582)) == 0.0
def test_cospatial_model():
# make the sampling deterministic
_ = np.random.seed(42)
source_size = UniformLikelihood.from_values("ss", 0.0, 400.0)
vlsrs = [
UniformLikelihood.from_values("vlsr", 0.0, 10.0) for _ in range(4)
]
ncols = [
UniformLikelihood.from_values("ncol", 0.0, 14.0) for _ in range(4)
]
Tex = GaussianLikelihood.from_values("tex", 5.8, 0.5, 0.0, 10.0)
dV = GaussianLikelihood.from_values("dV", 0.1, 1e-1, 0.0, 0.3)
model = CospatialTMC1(
source_size,
vlsrs,
ncols,
Tex,
dV,
None,
None,
)
# should be 14 parameters total; 2 * 4 components + 3
assert len(model) == 11
model._get_components()
# check the intialization routine is still working deterministically when asked
initial = model.sample_prior()
static = [
244.7411578889518,
9.50714306409916,
0.5808361216819946,
9.699098521619943,
3.0424224295953772,
10.247915185359671,
12.126466040849092,
11.654196971205904,
7.3465900428513295,
4.937541083743484,
0.043771247075902735,
]
assert np.allclose(initial, static)
# make sure the combined likelihood looks reasonable
parameters = np.asarray([
100.0,
5.6,
5.8,
5.95,
6.03,
10.24,
12.125,
11.65,
7.347,
5.87,
0.09216,
])
likelihood = model.compute_prior_likelihood(parameters)
assert np.round(abs(likelihood - 4.579927708578582)) == 0.0