def test_kalman_smoother_high_order_ibm():
"""The highest feasible order (without damping, which we dont use) is 11.
If this test breaks, someone played with the stable square-root
implementations in discrete_transition: for instance,
solve_triangular() and cho_solve() must not be changed to inv()!
"""
regression_problem, statespace_components = filtsmooth_zoo.car_tracking(
model_ordint=11,
timespan=(0.0, 1e-3),
step=1e-5,
forward_implementation="sqrt",
backward_implementation="sqrt",
)
truth = regression_problem.solution
kalman = filtsmooth.Kalman(
statespace_components["dynamics_model"],
statespace_components["measurement_model"],
statespace_components["initrv"],
)
posterior, _ = kalman.filtsmooth(regression_problem)
filtms = posterior.filtering_posterior.states.mean
smooms = posterior.states.mean
filtms_rmse = np.mean(np.abs(filtms[:, :2] - truth[:, :2]))
smooms_rmse = np.mean(np.abs(smooms[:, :2] - truth[:, :2]))
obs_rmse = np.mean(np.abs(regression_problem.observations - truth[:, :2]))
assert smooms_rmse < filtms_rmse < obs_rmse
def test_exactness_linear_model(self, rng):
"""Applied to a linear model, the results should be unchanged."""
# pylint: disable=not-callable
regression_problem, info = filtsmooth_zoo.car_tracking(rng=rng)
linear_model = info["prior_process"].transition
initrv = info["prior_process"].initrv
linearised_model = self.linearizing_component(linear_model)
# Assert that the objects are different
assert not isinstance(linear_model, type(linearised_model))
# Assert that the give the same outputs.
received, info1 = linear_model.forward_rv(initrv, 0.0)
expected, info2 = linearised_model.forward_rv(initrv, 0.0)
crosscov1 = info1["crosscov"]
crosscov2 = info2["crosscov"]
rtol, atol = 1e-9, 1e-9
np.testing.assert_allclose(received.mean,
expected.mean,
rtol=rtol,
atol=atol)
np.testing.assert_allclose(received.cov,
expected.cov,
rtol=rtol,
atol=atol)
np.testing.assert_allclose(crosscov1, crosscov2, rtol=rtol, atol=atol)
def test_kalman_smoother_high_order_ibm(rng):
"""The highest feasible order (without damping, which we dont use) is 11.
If this test breaks, someone played with the stable square-root implementations in
discrete_transition: for instance, solve_triangular() and cho_solve() must not be
changed to inv()!
"""
regression_problem, info = filtsmooth_zoo.car_tracking(
rng=rng,
num_prior_derivatives=11,
timespan=(0.0, 1e-3),
step=1e-5,
forward_implementation="sqrt",
backward_implementation="sqrt",
)
truth = regression_problem.solution
kalman = filtsmooth.gaussian.Kalman(info["prior_process"])
posterior, _ = kalman.filtsmooth(regression_problem)
filtms = posterior.filtering_posterior.states.mean
smooms = posterior.states.mean
filtms_rmse = np.mean(np.abs(filtms[:, :2] - truth[:, :2]))
smooms_rmse = np.mean(np.abs(smooms[:, :2] - truth[:, :2]))
obs_rmse = np.mean(np.abs(regression_problem.observations - truth[:, :2]))
assert smooms_rmse < filtms_rmse < obs_rmse
def test_exactness_linear_model(self):
"""Applied to a linear model, the results should be unchanged."""
# pylint: disable=not-callable
regression_problem, statespace_components = filtsmooth_zoo.car_tracking(
)
linear_model = statespace_components["dynamics_model"]
initrv = statespace_components["initrv"]
linearised_model = self.linearising_component_car(linear_model)
with self.subTest("Different objects"):
self.assertNotIsInstance(linear_model, type(linearised_model))
received, info1 = linear_model.forward_rv(initrv, 0.0)
expected, info2 = linearised_model.forward_rv(initrv, 0.0)
crosscov1 = info1["crosscov"]
crosscov2 = info2["crosscov"]
rtol, atol = 1e-10, 1e-10
with self.subTest("Same outputs"):
self.assertAllClose(received.mean,
expected.mean,
rtol=rtol,
atol=atol)
self.assertAllClose(received.cov,
expected.cov,
rtol=rtol,
atol=atol)
self.assertAllClose(crosscov1, crosscov2, rtol=rtol, atol=atol)
def test_kalman_value_error_repeating_timepoints(rng):
regression_problem, info = filtsmooth_zoo.car_tracking(
rng=rng,
num_prior_derivatives=4,
timespan=(0.0, 1e-3),
step=1e-5,
forward_implementation="sqrt",
backward_implementation="sqrt",
)
kalman = filtsmooth.gaussian.Kalman(info["prior_process"])
# This should raise a ValueError
regression_problem.locations[1] = regression_problem.locations[0]
with pytest.raises(ValueError):
posterior, _ = kalman.filtsmooth(regression_problem)
def test_kalman_multiple_measurement_models(rng):
regression_problem, info = filtsmooth_zoo.car_tracking(
rng=rng,
num_prior_derivatives=4,
timespan=(0.0, 1e-3),
step=1e-5,
forward_implementation="sqrt",
backward_implementation="sqrt",
)
truth = regression_problem.solution
kalman = filtsmooth.gaussian.Kalman(info["prior_process"])
posterior, _ = kalman.filtsmooth(regression_problem)
filtms = posterior.filtering_posterior.states.mean
smooms = posterior.states.mean
filtms_rmse = np.mean(np.abs(filtms[:, :2] - truth[:, :2]))
smooms_rmse = np.mean(np.abs(smooms[:, :2] - truth[:, :2]))
obs_rmse = np.mean(np.abs(regression_problem.observations - truth[:, :2]))
assert smooms_rmse < filtms_rmse < obs_rmse
def car_tracking3(rng):
return filtsmooth_zoo.car_tracking(
rng=rng, measurement_variance=1.23, timespan=(0.2, 10.2), step=0.5
)
def car_tracking2(rng):
return filtsmooth_zoo.car_tracking(
rng=rng, measurement_variance=0.5, timespan=(0.1, 10.1), step=0.5
)
def car_tracking1(rng):
return filtsmooth_zoo.car_tracking(
rng=rng, measurement_variance=2.0, timespan=(0.0, 10.0), step=0.5
)
def problem(rng):
"""Car-tracking problem."""
return filtsmooth_zoo.car_tracking(rng=rng)
import pytest
from probnum import problems
import probnum.problems.zoo.filtsmooth as filtsmooth_zoo
def rng():
return np.random.default_rng(seed=123)
# The logistic ODE problem is not random, so it does not get a random number generator.
all_filtmooth_setups = pytest.mark.parametrize(
"filtsmooth_setup",
[
filtsmooth_zoo.benes_daum(rng=rng()),
filtsmooth_zoo.car_tracking(rng=rng()),
filtsmooth_zoo.logistic_ode(),
filtsmooth_zoo.ornstein_uhlenbeck(rng=rng()),
filtsmooth_zoo.pendulum(rng=rng()),
],
)
@all_filtmooth_setups
def test_types(filtsmooth_setup):
regression_problem, info = filtsmooth_setup
assert isinstance(regression_problem, problems.TimeSeriesRegressionProblem)
assert isinstance(regression_problem.observations, np.ndarray)
assert isinstance(regression_problem.locations, np.ndarray)
assert isinstance(regression_problem.solution, np.ndarray)
assert isinstance(info, dict)
import numpy as np
import pytest
import probnum.problems.zoo.filtsmooth as filtsmooth_zoo
from probnum import problems
all_filtmooth_setups = pytest.mark.parametrize(
"filtsmooth_setup",
[
filtsmooth_zoo.benes_daum(),
filtsmooth_zoo.car_tracking(),
filtsmooth_zoo.logistic_ode(),
filtsmooth_zoo.ornstein_uhlenbeck(),
filtsmooth_zoo.pendulum(),
],
)
@all_filtmooth_setups
def test_types(filtsmooth_setup):
regression_problem, statespace_components = filtsmooth_setup
assert isinstance(regression_problem, problems.RegressionProblem)
assert isinstance(regression_problem.observations, np.ndarray)
assert isinstance(regression_problem.locations, np.ndarray)
assert isinstance(regression_problem.solution, np.ndarray)
assert isinstance(statespace_components, dict)
assert "dynamics_model" in statespace_components
assert "measurement_model" in statespace_components
assert "initrv" in statespace_components
