Source code for ewoksxas.fit.tests.test_lca_fit

import numpy as np
import pytest

from ewoksxas.fit.lca_fit import LCAModel

# The shared XYData container and as_1d_array helper (formerly duplicated here)
# are now covered by tests/test_data.py; this module tests LCAModel/FitResult.




def test_lca_fit_unbounded(sample, components):
    lca = LCAModel.from_dicts(sample, components)

    # Bounds
    assert lca.xmin == 0
    assert lca.xmax == np.pi
    assert lca.roi.x.min() == lca.xmin
    assert lca.roi.x.max() == lca.xmax
    assert lca.sample.x.size == lca.sample.y.size == 100
    assert lca.roi.x.size == lca.roi.y.size == 100

    # Components
    assert len(lca.components) == 5
    matrix, _ = lca._build_component_matrix()
    assert matrix.shape == (5, 100)

    # Fit
    fit = lca.fit()
    expected_solution = [0.1, 0.2, 0.3, 0.35, 0.05]
    assert np.allclose(expected_solution, fit.solution, atol=1e-6)
    assert np.allclose(fit.rmse, 0.0, atol=1e-6)
    assert np.allclose(fit.r2, 1.0, atol=1e-6)
    assert np.allclose(lca.roi.y, fit.model.y, atol=1e-6)





def test_lca_fit_bounded(sample, components):
    parameters = {"xmin": 1, "xmax": 2}
    lca = LCAModel.from_dicts(sample, components, parameters)

    # Bounds
    assert lca.xmin == 1
    assert lca.xmax == 2
    assert np.allclose([lca.roi.x.min(), lca.roi.x.max()], [1, 2], atol=0.1)
    assert lca.sample.x.size == lca.sample.y.size == 100
    assert lca.roi.x.size == lca.roi.y.size == 32

    # Components
    assert len(lca.components) == 5
    matrix, _ = lca._build_component_matrix()
    assert matrix.shape == (5, 32)

    # Fit
    fit = lca.fit()
    expected_solution = [0.1, 0.2, 0.3, 0.35, 0.05]
    assert np.allclose(expected_solution, fit.solution, atol=1e-3)
    assert np.allclose(fit.rmse, 0.0, atol=1e-6)
    assert np.allclose(fit.r2, 1.0, atol=1e-6)
    assert np.allclose(lca.roi.y, fit.model.y, atol=1e-6)





@pytest.mark.filterwarnings("ignore::RuntimeWarning")
def test_lca_fit_excess_components(sample, excess_components, rand_components):
    lca = LCAModel.from_dicts(sample, excess_components)
    fit = lca.fit()
    expected_solution = [0.1, 0.2, 0.3, 0.35, 0.05] + [0.0] * 9
    assert np.allclose(expected_solution, fit.solution, atol=1e-2)
    assert np.allclose(fit.rmse, 0.0, atol=1e-3)
    assert np.allclose(fit.r2, 1.0, atol=1e-3)
    assert np.allclose(lca.roi.y, fit.model.y, atol=1e-3)

    # randomize component list
    lca = LCAModel.from_dicts(sample, rand_components)
    fit = lca.fit()
    expected_solution = [
        0.35,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.3,
        0.0,
        0.05,
        0.0,
        0.1,
        0.2,
    ]
    assert np.allclose(expected_solution, fit.solution, atol=1e-2)
    assert np.allclose(fit.rmse, 0.0, atol=1e-3)
    assert np.allclose(fit.r2, 1.0, atol=1e-3)
    assert np.allclose(lca.roi.y, fit.model.y, atol=1e-3)