def _generate_observed_trajectories():
timepoints = np.arange(0, 20, 0.5)
observed_trajectories = [
Trajectory(timepoints, [
301., 290.1919552, 280.58701279, 272.03059275, 264.39179184,
257.55906225, 251.43680665, 245.94267027, 241.0053685,
236.56293031, 232.56126687, 228.95299703, 225.69647718,
222.75499485, 220.09609439, 217.69101017, 215.51418738,
213.54287512, 211.75677913, 210.13776407, 208.6695974,
207.33772795, 206.12909392, 205.03195581, 204.03575051,
203.13096347, 202.30901645, 201.5621687, 200.88342959,
200.26648137, 199.70561061, 199.19564728, 198.73191049,
198.3101601, 197.92655342, 197.57760651, 197.26015951,
196.97134545, 196.70856234, 196.46944793
], Moment([1], symbol='y_0')),
Trajectory(timepoints, [
0., 20.10320788, 35.54689328, 47.51901615, 56.88242563,
64.26983231, 70.14921364, 74.86943532, 78.69244584, 81.81623963,
84.39139622, 86.53309942, 88.32994353, 89.85043353, 91.1478162,
92.26369294, 93.23073689, 94.07474712, 94.81620873, 95.47148295,
96.05371852, 96.57355199, 97.03964777, 97.45911583, 97.83783557,
98.18070779, 98.49185119, 98.77475594, 99.03240415, 99.26736458,
99.48186728, 99.67786273, 99.85706877, 100.02100803, 100.17103799,
100.3083751, 100.43411443, 100.54924568, 100.65466632, 100.75119251
], Moment([2], symbol='yx1'))
]
return observed_trajectories
def test_trajectory_with_sensitivities_serialisation(self):
term = Moment([1, 0, 0], 'x')
x = Trajectory([1, 2, 3], [3, 2, 1], SensitivityTerm(term, 'x'))
y = Trajectory([1, 2, 3], [7, 8, 9], SensitivityTerm(term, 'y'))
t = TrajectoryWithSensitivityData([1, 2, 3], [-1, -2, -3],
term,
sensitivity_data=[x, y])
self._roundtrip(t)
def test_equality_treats_equal_things_as_equal(self):
"""
Given two Trajectories that were equal, they should be comparable with ==.
"""
t1 = Trajectory([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'))
t2 = Trajectory([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'))
self.assertEqual(t1, t2)
def test_single_traj_to_file(self):
trajectory = Trajectory([1, 2, 3, 4, 5, 6], [3, 2, 1,5, 2, 4], Moment([1], symbol='description'))
file = tempfile.mktemp(suffix=".csv")
try:
with open(file,"w") as out:
trajectory.to_csv(out)
finally:
os.unlink(file)
def test_different_descriptions_make_trajectories_different(self):
"""
Given two Trajectories that differ only by values, they should be treated as different
"""
t1 = Trajectory([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'))
t2 = Trajectory([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'))
self.assertNotEqual(t1, t2)
def test_equality_treats_equal_things_as_equal(self):
"""
Given two Trajectories that were equal, they should be comparable with ==.
"""
t_sensitivity_1 = Trajectory([1,2,3], [3, 2, 1], Moment([1], symbol='description'))
t_sensitivity_2 = Trajectory([1,2, 3], [-5, -9, -1], Moment([1], symbol='description'))
t1 = TrajectoryWithSensitivityData([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'), [t_sensitivity_1, t_sensitivity_2])
t2 = TrajectoryWithSensitivityData([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'), [t_sensitivity_1, t_sensitivity_2])
self.assertEqual(t1, t2)
def test_traj_collection_to_file(self):
tr2 = Trajectory([1, 2, 3, 4, 5, 6], [3, 2, 1, 5, 2, 4], Moment([1], symbol='y_1'))
tr1 = Trajectory([1, 2, 3, 4, 5, 6], [3, 2, 1, 5, 2, 4], Moment([1], symbol='y_2'))
tc = TrajectoryCollection([tr1,tr2])
file = tempfile.mktemp(suffix=".csv")
try:
with open(file,"w") as out:
tc.to_csv(out)
finally:
os.unlink(file)
def test_different_timepoints_make_trajectories_different(self):
"""
Given two TrajectoriesWithSensitivityData that differ only by sensitivity data
they should be reported as different
"""
t_sensitivity_1 = Trajectory([1,2,3], [3, 2, 1], Moment([1], symbol='sensitivity1'))
t_sensitivity_2 = Trajectory([1,2, 3], [-5, -9, -1], Moment([1], symbol='sensitivity2'))
t_sensitivity_3 = Trajectory([1,2,3], [-5, -9, -100], Moment([1], symbol='sensitivity2'))
t1 = TrajectoryWithSensitivityData([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'), [t_sensitivity_1, t_sensitivity_2])
t2 = TrajectoryWithSensitivityData([1, 2, 3], [3, 2, 1], Moment([1], symbol='description'), [t_sensitivity_1, t_sensitivity_3])
self.assertNotEqual(t1, t2)
def _test_trajectory_wrapping_works_correctly(self):
"""
Given a set of returned values from a solver, _results_to_trajectories should be able to
wrap those results into :class:`~means.simulation.trajectory.Trajectory` objects nicely.
"""
descriptions = [ODETermBase('x'), ODETermBase('y')]
simulated_timepoints = [1,2,3]
simulated_values = np.array([[-3, 1], [-2, 2], [-1, 3]])
trajectory1, trajectory2 = _wrap_results_to_trajectories(simulated_timepoints, simulated_values, descriptions)
correct_trajectory1 = Trajectory(simulated_timepoints, [-3, -2, -1], descriptions[0])
correct_trajectory2 = Trajectory(simulated_timepoints, [1, 2, 3], descriptions[1])
self.assertEqual(correct_trajectory1, trajectory1)
self.assertEqual(correct_trajectory2, trajectory2)
def _sample_inference():
r = Inference(problem=_sample_problem(),
starting_parameters=[1, 2, 3, 4, 5, 6, 7],
starting_conditions=[1, 2, 3],
variable_parameters=['c_0', 'c_1'],
observed_trajectories=[
Trajectory([1, 2], [2, 3], Moment([1, 0, 0], 'x'))
],
distance_function_type='gamma',
maxh=0.01) # Some simulation kwargs
return r
def _test_trajectory_with_sensitivity_wrapping_works_correctly(self):
"""
Given a set of already pre-processed trajectories, results from sensitivity calculations,
and the parameters of model, these should be wrapped nicely into TrajectoryWithSensitivityData objects
"""
time_points = [1, 2, 3]
t0 = Trajectory(time_points, [-1, 2, -3], ODETermBase('x'))
t1 = Trajectory(time_points, [-1, -2, -3], ODETermBase('y'))
parameters = sympy.symbols(['p', 'z'])
raw_sensitivity_results = np.array([[[np.nan, np.nan], [1, 2], [3, 4]],
[[np.nan, np.nan], [-1, -3], [-2, -4]]])
actual_t1, actual_t2 = _add_sensitivity_data_to_trajectories([t0, t1], raw_sensitivity_results, parameters)
sensitivity_data_1 = [Trajectory(time_points, [np.nan, 1, 3], SensitivityTerm(t0.description, parameters[0])),
Trajectory(time_points, [np.nan, 2, 4], SensitivityTerm(t1.description, parameters[1]))]
expected_t1 = TrajectoryWithSensitivityData.from_trajectory(t0, sensitivity_data_1)
sensitivity_data_2 = [Trajectory(time_points, [np.nan, -1, -2], SensitivityTerm(t0.description, parameters[0])),
Trajectory(time_points, [np.nan, -3, -4], SensitivityTerm(t1.description, parameters[1]))]
expected_t2 = TrajectoryWithSensitivityData.from_trajectory(t0, sensitivity_data_2)
self.assertEqual(actual_t1, expected_t1)
self.assertEqual(actual_t2, expected_t2)
def check_initialisation(variable_parameters,
expected_parameters_with_variability,
expected_initial_conditions_with_variability,
expected_constraints):
p = Inference(
self.dimer_problem, parameters, initial_conditions,
variable_parameters,
[Trajectory([1, 2, 3], [1, 2, 3], Moment([1], symbol='x'))])
self.assertEquals(p.starting_parameters_with_variability,
expected_parameters_with_variability)
self.assertEqual(p.starting_conditions_with_variability,
expected_initial_conditions_with_variability)
self.assertEqual(p.constraints, expected_constraints)
def test_trajectory_collection_serialisation(self):
t1 = Trajectory([1, 2, 3], [3, 2, 1], Moment([1, 2, 3], 'x'))
t2 = Trajectory([1, 2, 3], [3, 2, 1], Moment([1, 0, 0], 'y'))
tc = TrajectoryCollection([t1, t2])
self._roundtrip(tc)
def test_trajectory_serialisation(self):
t = Trajectory([1, 2, 3], [3, 2, 1], Moment([1, 2, 3], 'x'))
self._roundtrip(t)
def test_moment_from_traj(self):
dummy = StochasticProblem(MODEL_LOTKA_VOLTERRA)
ssa = SSASimulation(dummy, 1)
all_trajectories = [
[
Trajectory([0, 1, 2], [3, 4, 4], Moment([0,1],"a")),
Trajectory([0, 1, 2], [7, 9, 10], Moment([1,0],"b"))
],
[
Trajectory([0, 1, 2], [1, 2, 5], Moment([0,1],"a")),
Trajectory([0, 1, 2], [2, 3, 3], Moment([1,0],"b"))
],
[
Trajectory([0, 1, 2], [13,7,5], Moment([0,1],"a")),
Trajectory([0, 1, 2], [22,3,1], Moment([1,0],"b"))
],
[
Trajectory([0, 1, 2], [9,21, 3], Moment([0,1],"a")),
Trajectory([0, 1, 2], [8, 7, 4], Moment([1,0],"b"))
]
]
mean_trajectories = [sum(trajs)/float(len(trajs)) for trajs in zip(*all_trajectories)]
variance_a_result = ssa._compute_one_moment(all_trajectories, mean_trajectories,
Moment([2,0],sympy.Symbol("V_a")))
covar_result= ssa._compute_one_moment(all_trajectories, mean_trajectories,
Moment([1,1],sympy.Symbol("Cov_a_b")))
variance_a_expected = Trajectory([0.0, 1.0, 2.0],[22.75, 55.25, 0.6875],
Moment([2, 0], sympy.Symbol("V_a")))
covar_expected = Trajectory([0.0, 1.0, 2.0], [31.875, 5.75, -1.125],
Moment([1, 1], sympy.Symbol("Cov_a_b")))
skew_a_result = ssa._compute_one_moment(all_trajectories, mean_trajectories,
Moment([3,0],sympy.Symbol("V_a")))
skew_a_expected = Trajectory([0.0, 1.0, 2.0], [20.25, 396.0, -0.28125],
Moment([3, 0], symbol="V_a"))
self.assertEqual(variance_a_result , variance_a_expected)
self.assertEqual(covar_result, covar_expected)
self.assertEqual(skew_a_expected, skew_a_result)