def test_random_state():
# Testing that the `random_state` kwarg works as expected on the system
# Arbitrary wavelength wavelength points
wavelengths = np.linspace(8541, 8544, 49)
# Initialise model
model = IBIS8542Model(original_wavelengths=wavelengths, random_state=0)
# Get sample classifications
X, y = make_classification(200,
49,
n_classes=5,
n_informative=4,
random_state=0)
# Training #1
model.train(X[::2], y[::2])
score_a = cross_val_score(model.neural_network, X[1::2], y[1::2])
# Training #2
model.train(X[::2], y[::2])
score_b = cross_val_score(model.neural_network, X[1::2], y[1::2])
assert score_b == pytest.approx(score_a)
assert score_b == pytest.approx(np.array([0.45, 0.35, 0.45, 0.45, 0.35]))
def test_ibis8542model_basic():
# Will break if default parameters are changes in mcalf.models.IBIS8542Model
wl = 1000.97
x_orig = np.linspace(999.81, 1002.13, num=25)
prefilter_main = 1 - np.abs(x_orig - wl) * 0.1
prefilter_wvscl = x_orig - wl
m = IBIS8542Model(stationary_line_core=wl,
original_wavelengths=x_orig,
prefilter_ref_main=prefilter_main,
prefilter_ref_wvscl=prefilter_wvscl)
bg = 1327.243
arr = voigt(x_orig, -231.42, wl + 0.05, 0.2, 0.21, bg)
m.load_array(np.array([arr, arr]), names=['row', 'wavelength'])
m.load_background(np.array([bg, bg]), names=['row'])
# Fit with assumed neural network classification
fit1, fit2 = m.fit(row=[0, 1], classifications=np.array([0, 0]))
fit3 = m.fit_spectrum(arr, classifications=0,
background=bg) # Make sure this is equiv
truth = [-215.08275199, 1001.01035476, 0.00477063, 0.28869302]
assert np.array_equal(fit1.parameters, fit2.parameters)
assert np.array_equal(fit1.parameters, fit3.parameters)
assert list(fit1.parameters) == pytest.approx(truth)
def test_ibis8542model_validate_parameters(valid_kwargs):
stationary_line_core = valid_kwargs['stationary_line_core']
defaults = {
'absorption_guess': [-1000, stationary_line_core, 0.2, 0.1],
'emission_guess': [1000, stationary_line_core, 0.2, 0.1],
'absorption_min_bound':
[-np.inf, stationary_line_core - 0.15, 1e-6, 1e-6],
'emission_min_bound': [0, -np.inf, 1e-6, 1e-6],
'absorption_max_bound': [0, stationary_line_core + 0.15, 1, 1],
'emission_max_bound': [np.inf, np.inf, 1, 1],
'absorption_x_scale': [1500, 0.2, 0.3, 0.5],
'emission_x_scale': [1500, 0.2, 0.3, 0.5]
}
# Incorrect Lengths
for key, value in defaults.items(): # For each parameter
with pytest.raises(ValueError) as e: # Catch ValueErrors
defaults_mod = defaults.copy(
) # Create a copy of the default parameters
defaults_mod.update({key:
value[:-1]}) # Crop the parameter's value
IBIS8542Model(
**valid_kwargs, **defaults_mod
) # Pass the cropped parameter with the other default params
assert key in str(e.value) and 'length' in str(
e.value) # Error must be about length of current parameter
# Check that the sign of the following amplitudes are enforced
for sign, bad_number, parameters in [
('positive', -10.42, ('emission_guess', 'emission_min_bound')),
('negative', +10.42, ('absorption_guess', 'absorption_max_bound'))
]:
for p in parameters:
with pytest.raises(ValueError) as e:
bad_value = defaults[p].copy()
bad_value[0] = bad_number
defaults_mod = defaults.copy()
defaults_mod.update({p: bad_value})
IBIS8542Model(**valid_kwargs, **defaults_mod)
assert p in str(e.value) and sign in str(e.value)
def test_ibis8542model_configfile():
# Enter the data directory (needed to find the files referenced in the config files)
original_dir = os.getcwd()
os.chdir(data_path())
# Test with config file
IBIS8542Model(config="ibis8542model_config.yml")
# Turn off sigma
IBIS8542Model(config="ibis8542model_config.yml", sigma=False)
# Test with defined prefilter
with pytest.deprecated_call():
IBIS8542Model(config="ibis8542model_config_prefilter.yml")
# Test with no prefilter
IBIS8542Model(config="ibis8542model_config_noprefilter.yml")
# TODO Check that the parameters were imported correctly
# Go back to original directory
os.chdir(original_dir)
def ibis8542model_init():
# Enter the data directory (needed to find the files referenced in the config files)
original_dir = os.getcwd()
os.chdir(data_path())
x_orig = np.loadtxt("ibis8542model_wavelengths_original.csv")
m = IBIS8542Model(config="ibis8542model_init_config.yml",
constant_wavelengths=x_orig,
original_wavelengths=x_orig,
prefilter_response=np.ones(25))
m.neural_network = DummyClassifier(trained=True, n_features=25)
# Go back to original directory
os.chdir(original_dir)
return m
def test_ibis8542model_get_sigma():
# Enter the data directory (needed to find the files referenced in the config files)
original_dir = os.getcwd()
os.chdir(data_path())
m = IBIS8542Model(config="ibis8542model_config.yml")
sigma = np.loadtxt("ibis8542model_sigma.csv", delimiter=',')
assert np.array_equal(m._get_sigma(classification=0), sigma[0])
for c in (1, 2, 3, 4):
assert np.array_equal(m._get_sigma(classification=c), sigma[1])
for i in (0, 1):
assert np.array_equal(m._get_sigma(sigma=i), sigma[i])
x = np.array([1.4325, 1421.43, -1325.342, 153.3, 1.2, 433.0])
assert np.array_equal(m._get_sigma(sigma=x), x)
# Go back to original directory
os.chdir(original_dir)
wavelengths = np.delete(wavelengths, np.s_[1:6:2])
wavelengths = np.delete(wavelengths, np.s_[-6::2])
spectrum = double_voigt(wavelengths, -526, 8542, 0.1, 0.1,
300, 8541.9, 0.2, 0.05, 1242)
#%%
from mcalf.visualisation import plot_spectrum
plot_spectrum(wavelengths, spectrum, normalised=False)
#%%
# A basic model is created,
model = IBIS8542Model(original_wavelengths=wavelengths)
#%%
# The spectrum is provided to the model's fit method.
# A classifier has not been loaded so the classification must be
# provided manually.
# The fitting algorithm assumes that the intensity at the ends of the
# spectrum is zero, so in this case we need to provide it with a
# background value to subtract from the spectrum before fitting.
fit = model.fit(spectrum=spectrum, classifications=4, background=1242)
print(fit)
#%%
# The spectrum can now be plotted,