netcal.presentation.ReliabilityQCE

class netcal.presentation.ReliabilityQCE(bins: int | Iterable[float] | ndarray = 10)

Visualizes the Conditional Quantile Calibration Error (C-QCE) in the scope of regression calibration as a bar chart for probabilistic regression models. See netcal.metrics.regression.QCE for a detailed documentation of the C-QCE [1]. This method is able to visualize the C-QCE in terms of multiple univariate distributions if the input is given as multiple independent Gaussians. This method is also able to visualize the multivariate C-QCE for a multivariate Gaussian if the input is given with covariance matrices.

Parameters:

bins (int or iterable, default: 10) – Number of bins used by the C-QCE binning. If iterable, use different amount of bins for each dimension (nx1, nx2, … = bins).

References

[1]

Küppers, Fabian, Schneider, Jonas, and Haselhoff, Anselm: “Parametric and Multivariate Uncertainty Calibration for Regression and Object Detection.” European Conference on Computer Vision (ECCV) Workshops, 2022. Get source online

Methods

__init__([bins])	Constructor.
plot(X, y, q, *[, kind, range_, filename, ...])	Visualizes the C-QCE as a bar chart either for multiple univariate data (if standard deviations are given as input) or for a joint multivariate distribution (if covariance matrices are given as input).