pyuncertainnumber.validation.area_metric¶

Functions¶

`am_distance_counter`(A, B)	Compute the distance between two sets of intervals as used in the area metric.
`function_succeeds`(f, args, *kwargs)
`area_metric_pbox`(a, b)	when a and b are both Pboxes
`area_metric_pbox_diff`(a, b)	when a and b are both Pboxes
`area_metric_sample`(a, b)
`area_metric_np_numbers`(a, b)	when a and b are both numpy arrays of scalar numbers, compute the area metric accordingly
`area_metric_number`(→ float)	if any of a or b is a number, compute area metric accordingly
`area_metric`(→ float)	Compute the area metric between two objects.
`double_metric`(p, o)	Double metric for two uncertain numbers.
`conformal_double_metric`(p, o)	Propsed conformal version of the double metric, which takes the maximum of the two area metrics.
`am_diff_register`(A, B[, debug])	Register the distance and return a structured array with fields:
`intervals_from_res`(res)	From a structured array res with fields: 'distance', 'x1', 'x2',
`plot_intervals_from_res`(res, p, *[, show_points, ...])	Plot horizontal intervals from res at heights given by p.
`integrate_distance`(res, p)	Integrate distance vs p with the trapezoidal rule,

Module Contents¶

pyuncertainnumber.validation.area_metric.am_distance_counter(A, B)¶

Compute the distance between two sets of intervals as used in the area metric.

Notes

It is essentially doing:

def f(a, b, c, d):: return np.maximum.reduce([c - b, a - d, 0])

pyuncertainnumber.validation.area_metric.function_succeeds(f, *args, **kwargs)¶

pyuncertainnumber.validation.area_metric.area_metric_pbox(a: pyuncertainnumber.Pbox, b: pyuncertainnumber.Pbox)¶: when a and b are both Pboxes

pyuncertainnumber.validation.area_metric.area_metric_pbox_diff(a: pyuncertainnumber.Pbox, b: pyuncertainnumber.Pbox)¶: when a and b are both Pboxes

pyuncertainnumber.validation.area_metric.area_metric_sample(a: numpy.typing.ArrayLike, b: numpy.typing.ArrayLike)¶

pyuncertainnumber.validation.area_metric.area_metric_np_numbers(a, b)¶: when a and b are both numpy arrays of scalar numbers, compute the area metric accordingly

pyuncertainnumber.validation.area_metric.area_metric_number(a: pyuncertainnumber.Pbox | numbers.Number, b: pyuncertainnumber.Pbox | numbers.Number) → float¶: if any of a or b is a number, compute area metric accordingly

pyuncertainnumber.validation.area_metric.area_metric(a: numbers.Number | pyuncertainnumber.Pbox | numpy.typing.ArrayLike, b: numbers.Number | pyuncertainnumber.Pbox | numpy.typing.ArrayLike) → float¶: Compute the area metric between two objects.

Note

top-level function to compute area metric between any two objects

pyuncertainnumber.validation.area_metric.double_metric(p: numbers.Number | pyuncertainnumber.Pbox | pyuncertainnumber.Interval, o: numbers.Number | pyuncertainnumber.Pbox | pyuncertainnumber.Interval)¶

Double metric for two uncertain numbers.

Parameters:

p – a prediction uncertain number (Pbox)
o – an observation uncertain number (Pbox or scalar)

Note

Typical case is for validation between prediction and observation where both are uncertain numbers.

pyuncertainnumber.validation.area_metric.conformal_double_metric(p: numbers.Number | pyuncertainnumber.Pbox | pyuncertainnumber.Interval, o: numbers.Number | pyuncertainnumber.Pbox | pyuncertainnumber.Interval)¶: Propsed conformal version of the double metric, which takes the maximum of the two area metrics.

pyuncertainnumber.validation.area_metric.am_diff_register(A, B, debug=False)¶

Returns:

distance: float (max of c-b, a-d, 0)
x1: first endpoint (c or a, or 0 if 0 wins)
x2: second endpoint (b or d, or 0 if 0 wins)

Return type:

In each tuple of the output array

pyuncertainnumber.validation.area_metric.intervals_from_res(res)¶

From a structured array res with fields: ‘distance’, ‘x1’, ‘x2’, :returns: boolean mask where distance != 0

intervals: (N, 2) array of sorted [lo, hi] endpoints for rows where mask is True

Return type:

mask

pyuncertainnumber.validation.area_metric.plot_intervals_from_res(res, p, *, show_points=False, title=None, include_ties=False, ax=None)¶

Plot horizontal intervals from res at heights given by p.

args

res (ArrayLike) : Structured array with fields (‘distance’, ‘x1’, ‘x2’).

p (ArrayLike) : 1D array of same length as res, giving y-coordinate (height) of each interval.

show_points (Boolean): If True, shows markers at interval endpoints.

title (str): Plot title. Optional

include_ties (bool): If True, also include intervals where distance == 0 but argmax picked c-b or a-d.

ax (matplotlib.axes.Axes) : Axis to plot on. If None, a new figure and axis are created. Optional

returns

ax (matplotlib.axes.Axes): The matplotlib axis used for plotting.

pyuncertainnumber.validation.area_metric.integrate_distance(res, p)¶

Integrate distance vs p with the trapezoidal rule, matching: np.trapz(y=diff, x=p)

No masking; uses all rows. Sorts by p to avoid signed/ordering issues.