Skip to content

credibility.records

The data points of the adopted fits of a model with the kinds of their mappings.

The kind of a fit mapping (training, validation, outlier) decides what its data can be evidence of: a point which was fitted is no validation. The kinds are the ones the record of an adopted fit stores (metrics_mappings.tsv, the kind every mapping had when the fit ran), and the current selection of the fit definition must have the same kinds: a split changed after the adoption without a refit would otherwise let fitted data count as validation. The remedy of a difference is a refit and a new adoption, never an override.

evaluate_fits evaluates the adopted fits on the current model and data into the table of their data points; mapping_table is the same table with one row per mapping, without a simulation, for the static check of a declaration.

FitTable dataclass 

FitTable(
    *,
    table,
    lookup,
    recorded_kinds,
    reports=(lambda: MappingProxyType({}))(),
)

The data points of the adopted fits of a model.

Attributes:

Name Type Description
table DataFrame

one row per data point (evaluate_fits) or per mapping (mapping_table), with the columns fit (the subset of the fit), experiment, mapping, kind, observable and the fields of the metadata of the mappings; the data points have the columns of sbmlsim.fit.metrics.FitMetrics.datapoints_df as well.
lookup MappingLookup

the fit mapping of an experiment and a key.
recorded_kinds Kinds

the kinds the records of the adopted fits store, by the subset of the fit, the experiment and the key of the mapping.
reports Mapping[str, FitReport]

the reports the data points were evaluated with, by the subset of the fit (evaluate_fits), which the verification evaluates again with other settings; none for mapping_table.

KindMismatchError 

KindMismatchError(problems)

Bases: ValueError

The kinds of the fit mappings differ from the kinds the adopted fits recorded.

Attributes:

Name Type Description
problems

every mapping whose kind differs.

Create the error.

Parameters:

Name Type Description Default
problems list[str]

every mapping whose kind differs.

 required

recorded_kinds 

recorded_kinds(record, fit)

Get the kinds of the mappings which the record of an adopted fit stores.

Parameters:

Name Type Description Default
record Path

the directory of the record of the adopted fit.

 required
fit str

the subset of the fit, the first item of every key.

 required

Returns:

Type Description
dict[tuple[str, str, str], str]

The kind by the subset, the experiment and the key of the mapping.

Raises:

Type Description
FileNotFoundError

if the record has no metrics of its mappings.
ValueError

if the metrics lack the experiment, the mapping or the kind, or store two kinds of one mapping.

kind_differences 

kind_differences(table, recorded)

Compare the kinds of the mappings of a table with the recorded kinds.

Parameters:

Name Type Description Default
table DataFrame

a table of the fits with the columns fit, experiment, mapping and kind, see FitTable.

 required
recorded Kinds

the recorded kinds, see recorded_kinds.

 required

Returns:

Type Description
list[str]

One message per mapping whose kind differs from the recorded one, or
list[str]

which only the table or only the record has; empty if the kinds agree.

adopted_ids 

adopted_ids(project)

Get the ids of the adopted fits of a project, in the order of their subsets.

The subsets are the keys of the file of the adopted fits (Project.adopted_fits), so a subset has one adopted fit; the id of a fit must be one of its subset (sbmlsim.fit.result.fit_subset).

Raises:

Type Description
ValueError

if the project has no adopted fits or a fit id is no fit of the subset it is adopted for.

evaluate_fits 

evaluate_fits(project)

Evaluate the adopted fits of a project into the table of their data points.

Every fit is evaluated fresh on the current model and data with the settings it ran with (sbmlsim.fit.cli.report_from_records), at the reference parameter set of its record; the kinds of its mappings must be the recorded ones.

Parameters:

Name Type Description Default
project Project

the project, its adopted fits are evaluated.

 required

Returns:

Type Description
FitTable

The data points of every fit with the column fit, the subset, and the
FitTable

report of every fit.

Raises:

Type Description
ValueError

if the project has no adopted fits, see adopted_ids.
FileNotFoundError

if a record has no parameters or no kinds.
KindMismatchError

if a kind differs from the recorded one.

mapping_table 

mapping_table(project)

Get the mappings of the adopted fits of a project without a simulation.

The kinds are the ones of the current selection of every fit definition; the excluded mappings, which a fit does not evaluate, are no rows. The recorded kinds are compared by the caller, see kind_differences.

Parameters:

Name Type Description Default
project Project

the project, its adopted fits are listed.

 required

Returns:

Type Description
FitTable

One row per fit and mapping with the columns of KEY_COLUMNS and the
FitTable

fields of the metadata.

Raises:

Type Description
ValueError

if the project has no adopted fits, see adopted_ids.
FileNotFoundError

if a record has no parameters or no kinds.