eqiora.time

← Python API index

Time-integration methods and related settings.

View source

Example: control ODE error

Use model and x from the decay example:

python

import eqiora

method = eqiora.time.Tsitouras45(
    initial_step_s=0.01,
    relative_tolerance=1e-9,
    absolute_tolerances={(x, 0, 0, False): 1e-11},
)
plan = eqiora.resolve(model, temporal=method)

The initial step is in seconds. Each absolute tolerance applies to one state coordinate (field, derivative_order, component, imaginary). Components use zero-based row-major indices; False selects the real part and True the imaginary part. Tolerances use the coordinate’s SI units; x is dimensionless here. output_times_s in run chooses observation times independently of the adaptive internal steps.

API

EulerMaruyama

Ito diffusion on 2**depth equal steps over the original Run horizon.

Output times must lie on that dyadic grid. Exact restart retains the original partition; this profile does not promise cross-platform random bytes.

python

@final
class EulerMaruyama:
    def __new__(cls, *, depth: int) -> Self: ...
    @property
    def depth(self) -> int: ...
    def __repr__(self) -> str: ...

PoissonEvents

Constant-rate independent Poisson arrivals with a total event budget.

python

@final
class PoissonEvents:
    def __new__(cls, *, max_events: int) -> Self: ...
    @property
    def max_events(self) -> int: ...
    def __repr__(self) -> str: ...

BackwardEuler

Backward Euler with a fixed positive step size.

python

@final
class BackwardEuler:
    def __new__(cls, step_s: float) -> Self: ...
    @property
    def step_s(self) -> float: ...
    def __repr__(self) -> str: ...

TimePlanView

Resolved plan for finite time evolution (no mesh).

python

@final
class TimePlanView:
    @property
    def kind(self) -> str: ...
    @property
    def backend(self) -> str: ...
    @property
    def backend_version(self) -> str: ...
    def __repr__(self) -> str: ...
    @property
    def scalar_type(self) -> str: ...
    @property
    def value_representation(self) -> str: ...

Tsitouras45

Adaptive explicit Tsitouras 5(4) integration.

Tolerances are in SI units and are set per Field, derivative order, component, and real or imaginary part.

python

@final
class Tsitouras45:
    def with_conserved_norm(self, fields: Sequence[FieldRef], *, target: float, tolerance: float, dimension: Dimension) -> Self: ...
    def with_hermitian_parameter(self, parameter: ParameterRef) -> Self: ...
    def __new__(cls, *, initial_step_s: float, relative_tolerance: float, absolute_tolerances: Mapping[tuple[FieldRef, int, int, bool], float], events: EventPolicy | None=None, forward_sensitivities: ForwardSensitivity | None=None) -> Self: ...
    @property
    def initial_step_s(self) -> float: ...
    @property
    def relative_tolerance(self) -> float: ...
    @property
    def absolute_tolerances(self) -> dict[tuple[FieldRef, int, int, bool], float]: ...
    @property
    def events(self) -> EventPolicy | None: ...
    @property
    def forward_sensitivities(self) -> ForwardSensitivity | None: ...
    def __repr__(self) -> str: ...

ImplicitMidpoint

Implicit midpoint rule with a fixed step and per-coordinate Newton tolerances.

The tolerances control the Newton corrections, not the time-discretization error. Output samples are interpolated from the collocation polynomial of each completed step, so requesting outputs does not change the steps.

python

@final
class ImplicitMidpoint:
    def with_conserved_norm(self, fields: Sequence[FieldRef], *, target: float, tolerance: float, dimension: Dimension) -> Self: ...
    def with_hermitian_parameter(self, parameter: ParameterRef) -> Self: ...
    def __new__(cls, *, step_s: float, relative_tolerance: float, absolute_tolerances: Mapping[tuple[FieldRef, int, int, bool], float]) -> Self: ...
    @property
    def step_s(self) -> float: ...
    @property
    def relative_tolerance(self) -> float: ...
    @property
    def absolute_tolerances(self) -> dict[tuple[FieldRef, int, int, bool], float]: ...
    def __repr__(self) -> str: ...

TimeObservationInterpolation

Interpolation of Observable samples for time integrals and extrema.

EndpointLinear uses accepted endpoints. NativeMidpointQuadratic also uses native accepted midpoint values; it is unavailable for spatial histories. Neither policy interpolates a restart State.

python

@final
class TimeObservationInterpolation:
    NativeMidpointQuadratic: ClassVar[TimeObservationInterpolation]
    EndpointLinear: ClassVar[TimeObservationInterpolation]
    def __eq__(self, other: object, /) -> bool: ...
    def __hash__(self) -> int: ...

GuardTolerance

Positive tolerance, in coherent SI units, for one activation guard.

python

@final
class GuardTolerance:
    def __new__(cls, activation: ActivationRef, value: float, dimension: Dimension) -> Self: ...
    @property
    def activation(self) -> ActivationRef: ...
    @property
    def value(self) -> float: ...
    @property
    def dimension(self) -> Dimension: ...

EventPolicy

Event-handling settings, with guard tolerances in each Activation’s units.

Events that must fire together as one root group need identical guard tolerances.

python

@final
class EventPolicy:
    def __new__(cls, *, max_events: int, guard_tolerances: tuple[GuardTolerance, ...]) -> Self: ...
    @property
    def max_events(self) -> int: ...
    @property
    def model_digest(self) -> str: ...
    @property
    def guard_tolerances(self) -> tuple[GuardTolerance, ...]: ...

ForwardSensitivity

Error control for continuous forward sensitivities.

python

@final
class ForwardSensitivity:
    def __new__(cls, *, relative_tolerance: float, absolute_tolerances: tuple[SensitivityTolerance, ...]) -> Self: ...
    @property
    def relative_tolerance(self) -> float: ...
    @property
    def model_digest(self) -> str: ...
    @property
    def absolute_tolerances(self) -> tuple[SensitivityTolerance, ...]: ...

SensitivityTolerance

Positive tolerance for the derivative of a Field with respect to a Parameter, in the units of that derivative.

python

@final
class SensitivityTolerance:
    def __new__(cls, coordinate: tuple[FieldRef, int, int, bool], parameter: ParameterRef, value: float, dimension: Dimension) -> Self: ...
    @property
    def coordinate(self) -> tuple[FieldRef, int, int, bool]: ...
    @property
    def parameter(self) -> ParameterRef: ...
    @property
    def value(self) -> float: ...
    @property
    def dimension(self) -> Dimension: ...