Visualisations

Visualisations file describes all visualisations tools in PyCalib

pycalib.visualisations.plot_reliability_diagram(labels, scores, legend=None, show_histogram=True, bins=10, class_names=None, fig=None, show_counts=False, errorbar_interval=None, interval_method='beta', fmt='s-', show_correction=False, show_gaps=False, sample_proportion=0, hist_per_class=False, color_list=None, show_bars=False, invert_histogram=False, color_gaps='lightcoral', confidence=False, ax=None)

Plots the reliability diagram of the given scores and true labels

Parameters:

labelsarray (n_samples, ): Labels indicating the true class.
scoresmatrix (n_samples, n_classes) or list of matrices: Output probability scores for one or several methods.
legendlist of strings or None: Text to use for the legend.
show_histogramboolean: If True, it generates an additional figure showing the number of samples in each bin.
binsint or list of floats: Number of bins to create in the scores’ space, or list of bin boundaries.
class_nameslist of strings or None: Name of each class, if None it will assign integer numbers starting with 1.
figmatplotlib.pyplot.Figure or None: Figure to use for the plots, if None a new figure is created.
show_countsboolean: If True shows the number of samples of each bin in its corresponding line marker.
errorbar_intervalfloat or None: If a float between 0 and 1 is passed, it shows an errorbar corresponding to a confidence interval containing the specified percentile of the data.
interval_methodstring (default: ‘beta’): Method to estimate the confidence interval which uses the function proportion_confint from statsmodels.stats.proportion
fmtstring (default: ‘s-‘): Format of the lines following the matplotlib.pyplot.plot standard.
show_correctionboolean: If True shows an arrow for each bin indicating the necessary correction to the average scores in order to be perfectly calibrated.
show_gapsboolean: If True shows the gap between the average predictions and the true proportion of positive samples.
sample_proportionfloat in the interval [0, 1] (default 0): If bigger than 0, it shows the labels of the specified proportion of samples.
hist_per_classboolean: If True shows one histogram of the bins per class.
color_listlist of strings or None: List of string colors indicating the color of each method.
show_barsboolean: If True shows bars instead of lines.
invert_histogramboolean: If True shows the histogram with the zero on top and highest number of bin samples at the bottom.
color_gapsstring: Color of the gaps (if shown).
confidenceboolean: If True shows only the confidence reliability diagram.

Returns:

figmatplotlib.pyplot.figure: Figure with the reliability diagram