Skip to content

Instance Based

CF(mlmodel, mode)

Bases: InstanceBase

Abstract class to implement Coutnterfactual Methods for time series data

Initialization of CF. Arguments: mlmodel [torch.nn.Module, Callabale, tf.keras.Model]: Machine Learning Model to be explained. mode str : Second dimension is feature --> 'feat', is time --> 'time'

explain()

Explains instance or model.

plot(original, org_label, exp, exp_label, vis_change=True, all_in_one=False, save_fig=None, figsize=(6.4, 4.8))

Basic Plot Function for visualizing Coutnerfactuals. Arguments: original np.array: Instance to be explained. Shape: mode = time -> (time, feat) or mode = time -> (feat, time) org_label int: Label of instance to be explained. exp np.array: Explanation. mode = time -> (time, feat) or mode = time -> (feat, time) exp_label int: Label of Explanation. vis_change bool: Change to be visualized as heatmap. all_in_one bool: Original and Counterfactual in one plot. save_fig str: Path to save fig at.

plot_in_one(item, org_label, exp, cf_label, save_fig=None, figsize=(6.4, 4.8))

Plot Function for Counterfactuals in uni-and multivariate setting. In the multivariate setting only the changed features are visualized. Arguments: item np.array: original instance. Shape: mode = time -> (time, feat) or mode = time -> (feat, time) org_label int: originally predicted label. exp np.array: returned explanation. Shape: mode = time -> (time, feat) or mode = time -> (feat, time) cf_label int: lebel of returned instance. figsize Tuple: Size of Figure (x,y). save_fig str: Path to Save the figure.

plot_multi(item, org_label, exp, cf_label, figsize=(6.4, 4.8), save_fig=None)

Plot Function for Ates et al., used if multiple features are changed in a Multivariate Setting. Also called via plot_in_one. Preferably, do not use directly. Arguments: item np.array: original instance. Shape: mode = time -> (time, feat) or mode = time -> (feat, time) org_label int: originally predicted label. exp np.array: returned explanation. Shape: mode = time -> (time, feat) or mode = time -> (feat, time) cf_label int: lebel of returned instance. figsize Tuple: Size of Figure (x,y). save_fig str: Path to Save the figure.