beexai.evaluate package

Subpackages

Submodules

beexai.evaluate.comparison module

beexai.evaluate.plot_metric module

Radar plot for explanation metrics

class beexai.evaluate.plot_metric.ComplexRadar(fig, variables, ranges, n_ordinate_levels=6)[source]

Bases: object

fill(data, *args, **kw)[source]
fill_multiple(data, *args, **kw)[source]
plot(data, *args, **kw)[source]
plot_multiple(data, methods, *args, **kw)[source]
beexai.evaluate.plot_metric.get_dec_exponent(x)[source]

Get the decimal exponent of a number x

beexai.evaluate.plot_metric.plot_metric(df_path, metrics_plot=['FaithCorr_1-', 'Sensitivity_0+', 'Infidelity_0+', 'Comprehensiveness_1-', 'Sufficiency_0+', 'AUC_TP_0+', 'Monotonicity_1-', 'Complexity_0+', 'Sparseness_1-'], methods_plot=['Lime', 'ShapleyValueSampling', 'KernelShap', 'DeepLift', 'IntegratedGradients', 'Saliency'], plot_nn=True, save_path=None, alpha=0.2) → None[source]

Plot the radar chart for the metrics.

Parameters:

  • df_path – path for the metrics dataframe

  • metrics_plot – list of metrics to plot

  • methods_plot – list of methods to plot

  • plot_nn – whether to plot metric for Neural Network

  • save_path – path to save the plot

  • alpha – transparency of the plot

beexai.evaluate.switching_point module

Module contents