TSEvo

`TSEvo(model, data, transformer='authentic_opposing_information', epochs=500, mode='time', backend='PYT', verbose=0)`

Bases: CF

Calculates and Visualizes Counterfactuals for Uni- and Multivariate Time Series in accordance to the paper [1].

References

[1] Höllig, Jacqueline , et al. "TSEvo: Evolutionary Counterfactual Explanations for Time Series Classification." 21st IEEE International Conference on Machine Learning and Applications (ICMLA). IEEE, 2022.

PARAMETER	DESCRIPTION
`model`	Model to be interpreted. TYPE: `[Module, Callable, model]`
`data`	Reference Dataset as Tuple (x,y). TYPE: `Tuple`
`transformer`	['authentic_opposing_information','mutate_both','mutate_mean','frequency_band_mapping'] TYPE: `str` DEFAULT: `'authentic_opposing_information'`
`epochs`	Maximal Number of Itertions TYPE: `int` DEFAULT: `500`
`mode`	Name of second dimension: time -> (-1, time, feature) or feat -> (-1, feature, time) TYPE: `str` DEFAULT: `'time'`
`backend`	desired Model Backend ('PYT', 'TF', 'SK'). TYPE: `str` DEFAULT: `'PYT'`
`verbose`	Logging Level TYPE: `int` DEFAULT: `0`

`explain(original_x, original_y, target_y=None)`

Entry Point to explain a instance. Arguments: original_x (np.array): The instance to explain. Shape mode = time -> (1,time, feat) or mode = time -> (1,feat, time) original_y (np.array): Classification Probability of instance. target_y int: Class to be targeted

RETURNS	DESCRIPTION
	[np.array, int]: Returns the Counterfactual and the class. Shape of CF : `mode = time` -> `(time, feat)` or `mode = time` -> `(feat, time)`