fit.petab_v2.sciml_export¶

Write the neural networks of a hybrid problem as PEtab SciML.

The inverse of sbmlsim.fit.petab_v2.sciml: the hybridizations of an OptimizationProblem become the block sciml of the problem, the hybridization table, the rows of the mapping table and of the parameter table which name the networks, the NN YAML of every network and one array file per network with its arrays and the arrays of its inputs.

`sbmlsim`	PEtab SciML
`Network`	the NN YAML and the arrays of the array file
`Hybridization.pattern`	`pre_initialization` of the network, and the observables
an input which is a formula	a row of the hybridization table, evaluated along the trajectory
an input which is a parameter of its own	the parameter is the `petabEntityId` of the input
an input with a formula per condition	a change of the condition of every experiment
the formula of the other conditions	a change of the conditions without a formula of their own, and the block `sbmlsim`
an input which is arrays	`array` in the hybridization table, the arrays in the array file by condition
an output of `RHS` or `PRE_INITIALIZATION`	a row of the hybridization table which assigns the target
an output of `OBSERVABLE`	the symbol of the observable formula, mapped to the output
`frozen` and the bounds of the elements	the most general rows of the parameter table
`constants`	rows of the parameter table which are not estimated
a parameter of the model an input before the simulation or per condition uses	a row of the parameter table which is not estimated

A formula which holds for every condition is a row of the hybridization table, which PEtab SciML evaluates along the trajectory; a condition sets a value once at the start of a period and uses only the parameters of the parameter table, so only the inputs with a formula per condition are changes of the conditions. The linter of petab counts a parameter of the parameter table which the model does not have, i.e. a constant of a hybridization or an estimated parameter which is no entity of the model, as used where the mapping table names it or a condition uses it, not where the hybridization table uses it: an input whose formula is such a parameter, which no other formula uses, is the parameter (its petabEntityId, as the cases of PEtab SciML write it), and such a parameter in a row of the hybridization table gets a row of the mapping table without a modelEntityId, an annotation. An input before the simulation and a condition use only parameters of the table: a parameter of the model which the fit does not estimate is written as a row which is not estimated, any other entity of the model is refused (gap sciml-input-formula). The formula of an input for the conditions without a formula of their own is written into the conditions of those experiments and into the block sbmlsim, from which the reader gives the input back as it was.

The module imports sbmlsim.sciml and with it petab_sciml, which is the extra sciml. sbmlsim.fit.petab_v2.export imports it only for a problem with hybridizations.

ArrayRow `dataclass` ¶

ArrayRow(estimate, lower, upper)

The row of the parameter table an array of a network is described by.

key ¶

key()

Get the row as a hashable key.

ExportedNetwork `dataclass` ¶

ExportedNetwork(
    network,
    hybridizations,
    pre_initialization,
    arrays,
    parameter_rows=dict(),
)

A network of the problem with what is written for it.

Attributes:

Name	Type	Description
`network`	`Network`	the network.
`hybridizations`	`list[Hybridization]`	its hybridizations, one or two.
`pre_initialization`	`bool`	whether it runs before the simulation.
`arrays`	`ArrayData`	the array file of the network, with the arrays of its inputs.
`parameter_rows`	`dict[str, tuple[str, ArrayRow]]`	the rows of the parameter table, `petabEntityId` -> `modelEntityId` and the row.

SciMLExporter ¶

SciMLExporter(
    problem,
    hybridizations,
    simulation_ids,
    parameter_set=None,
)

Write the networks of an optimization problem as PEtab SciML.

The exporter is created by PetabExporter for a problem with hybridizations and fills the tables of the problem it builds. It validates the hybridizations against the fit parameters of the problem: every element of a network which is a fit parameter has the value of the network as its start value, the linear scale and the unit dimensionless, and the elements of one array are either all estimated with one pair of bounds or all frozen.

Attributes:

Name	Type	Description
`problem`		the problem which is exported.
`simulation_ids`		id of the simulation of a fit mapping -> ids of its PEtab experiments.
`networks`	`dict[str, ExportedNetwork]`	the networks by their id.
`element_ids`	`set[str]`	the ids of the elements of all networks, which are no rows of the parameter table and no part of the `sbmlsim` block.
`constants`	`dict[str, float]`	the constants of the hybridizations which are rows of the parameter table, id -> value.
`input_ids`	`dict[str, str]`	id of an input -> its `petabEntityId`, which is the parameter of an input whose formula is a parameter of its own and the id of the input otherwise.
`input_formulas`	`dict[str, dict[str, str]]`	id of an input which is written as changes of the conditions -> id of the simulation -> the formula.
`fallbacks`	`dict[str, str]`	id of an input -> its formula for the simulations without a formula of their own, in the math of PEtab, which the conditions repeat and the block `sbmlsim` carries.
`row_inputs`	`dict[str, str]`	id of an input which is a row of the hybridization table -> its formula.
`annotations`	`dict[str, str]`	id of a parameter which only rows of the hybridization table use -> the first input which uses it, a row of the mapping table without a `modelEntityId`.
`fixed`	`dict[str, float]`	id of a parameter of the model which an input before the simulation or a condition uses and the fit does not estimate -> its value, a row of the parameter table which is not estimated.
`hybridization_rows`	`list[Hybridization]`	the rows of the hybridization table.

Initialize the exporter.

Parameters:

Name	Type	Description	Default
`problem`	`OptimizationProblem`	the initialized problem.	required
`hybridizations`	`Sequence[Any]`	the hybridizations of the problem.	required
`simulation_ids`	`Mapping[str, Sequence[str]]`	id of the simulation of a fit mapping (the condition of the inputs) -> ids of its experiments of PEtab: a simulation whose fit mappings are in several collections is several experiments, which have the same inputs.	required
`parameter_set`	`ParameterSet \| None`	values of the parameters of the fit, e.g. the result of a fit. The arrays of the networks are written with the values of the elements in the set instead of the values the networks have, the elements which are not in the set keep theirs; a value for a frozen element is refused.	`None`

Raises:

Type Description

ValueError

if a hybridization is not a Hybridization of sbmlsim.sciml, if a network is before the simulation and in the model, if two hybridizations of a network differ in the network, the model or the inputs, if an element which is a fit parameter differs from the network, if the elements of an array are not described by one row, if two hybridizations give a constant different values, if an input has no formula for a simulation of the problem, or if the parameter set has a value for an element which is no parameter of the fit.

place_inputs ¶

place_inputs(models, model_changes)

Decide where every input of a formula is written, see the module.

Fills input_ids, input_formulas, fallbacks, row_inputs, annotations, fixed and hybridization_rows; PetabExporter calls it once the models of the problem are added.

Parameters:

Name	Type	Description	Default
`models`	`Mapping[str, Any]`	id of a model of the problem -> its `SbmlModel` of `petab`, the model without the networks.	required
`model_changes`	`Mapping[str, Mapping[str, float]]`	id of a model -> the changes the fit applies to it, id -> value.	required

Raises:

Type	Description
`ValueError`	if an input has no formula for a simulation of the problem, if the model of a network is not a model of the problem, or if an input which a condition sets or which runs before the simulation uses an entity of the model which is no parameter of the parameter table (gap 'sciml-input-formula').

conditions_of ¶

conditions_of(simulation)

Get the ids of the conditions of the first periods of a simulation.

Parameters:

Name	Type	Description	Default
`simulation`	`str`	id of the simulation.	required

Returns:

Type	Description
`list[str]`	The id of the condition of the first period of every experiment of
`list[str]`	the simulation, which `PetabExporter._periods` gives it; none for
`list[str]`	a simulation none of whose fit mappings is written.

input_changes ¶

input_changes(simulation)

Get the changes of the condition of a simulation which set inputs.

Parameters:

Name	Type	Description	Default
`simulation`	`str`	id of the simulation.	required

Returns:

Type	Description
`list[Change]`	One change per input which is written as changes of the
`list[Change]`	conditions, see `place_inputs`.

needs_condition ¶

needs_condition()

Check whether the first period of the experiments needs a condition.

An input which differs between the conditions needs the condition of every experiment: its formula is a change of it, its arrays are keyed by it in the array file.

mapping_rows ¶

mapping_rows()

Get the rows of the mapping table which name parts of the networks.

Returns:

Type	Description
`list[Mapping]`	A row per input, per used output and per row of the parameter
`list[Mapping]`	table of the networks. The `petabEntityId` of an input and of an
`list[Mapping]`	output which sets an entity is its id in `sbmlsim`, the one of an
`list[Mapping]`	input which is a constant of its own is the constant, the one of
`list[Mapping]`	an output an observable uses is its target, i.e. the symbol of
`list[Mapping]`	the observable formula.

parameter_rows ¶

parameter_rows()

Get the rows of the parameter table of the networks and the constants.

Returns:

Type	Description
`list[Parameter]`	The rows of the arrays of every network with the nominal value
`list[Parameter]`	`array`, and one row per constant of the hybridizations which is
`list[Parameter]`	not estimated.

config ¶

config()

Get the block of the extension sciml of the problem.

write ¶

write(output_dir)

Write the files of the networks into the directory of the problem.

Parameters:

Name	Type	Description	Default
`output_dir`	`Path`	the directory of the problem. The NN YAML and the array file of every network and the hybridization table are written next to `problem.yaml`.	required

network_yaml ¶

network_yaml(sid)

Get the file of the NN YAML of a network.

network_arrays ¶

network_arrays(sid)

Get the array file of a network, with its arrays and its inputs.

petab_math ¶

petab_math(formula)

Get the math of PEtab of an L3 formula of SBML.

Parameters:

Name	Type	Description	Default
`formula`	`str`	the formula, e.g. `alpha + (prey - 1.3)`.	required

Returns:

Type	Description
`str`	The expression in the math of PEtab, in which `log` is the natural
`str`	logarithm. A decimal of the formula is written as a decimal, not as
`str`	the fraction the parser reads it as.

Raises:

Type	Description
`ValueError`	if the formula is not valid math.

petab_index ¶

petab_index(index, shape)

Get the index of an element of an output as PEtab SciML writes it.

Parameters:

Name	Type	Description	Default
`index`	`tuple[int, ...]`	the index of the element in the output.	required
`shape`	`tuple[int, ...]`	the shape of the output.	required

Returns:

Type	Description
`str`	The index, e.g. `[0][1]`, without the leading axes of length one,
`str`	which are the axes of the batch, see `sbmlsim.fit.petab_v2.sciml`.

parameters_id ¶

parameters_id(network, layer=None, array=None)

Get the petabEntityId of the parameters of a network, a layer or an array.

model_entity_id ¶

model_entity_id(network, layer=None, array=None)

Get the modelEntityId of the parameters of a network, a layer or an array.

fit.petab_v2.sciml_export¶

ArrayRow dataclass ¶

key ¶

ExportedNetwork dataclass ¶

SciMLExporter ¶

place_inputs ¶

conditions_of ¶

input_changes ¶

needs_condition ¶

mapping_rows ¶

parameter_rows ¶

config ¶

write ¶

network_yaml ¶

network_arrays ¶

petab_math ¶

petab_index ¶

parameters_id ¶

model_entity_id ¶

ArrayRow `dataclass` ¶

ExportedNetwork `dataclass` ¶