peal.explainers.no_generator_counterfactual_explainers.DiCEExplainerConfig¶

class peal.explainers.no_generator_counterfactual_explainers.DiCEExplainerConfig[source]¶

Config of DiCEExplainer.

Parameters:

predictor_path (str, optional) – Path of the predictor to explain when run stand-alone.
distilled_predictor (str or dict, optional) – Optional student predictor spec (kept for API parity with CFKD).
y_target_goal_confidence (float) – Target-class confidence regarded as a successful flip.
num_attempts (int) – Number of counterfactuals DiCE is asked for per factual (total_CFs).
parallel_attempts (int) – Kept for API parity; DiCE explains samples one at a time.
method (str) – DiCE search method: "random", "genetic" or "kdtree".
features_to_vary (list or str) – Feature names DiCE may change, or "all".
proximity_weight (float) – DiCE loss weights. Stored on the config but not currently passed to generate_counterfactuals.
sparsity_weight (float) – DiCE loss weights. Stored on the config but not currently passed to generate_counterfactuals.
diversity_weight (float) – DiCE loss weights. Stored on the config but not currently passed to generate_counterfactuals.

Methods

Attributes

`distilled_predictor`
`diversity_weight`
`explainer_type`	The type of explanation that shall be used.
`features_to_vary`
`method`
`num_attempts`	The number of counterfactuals created for the same sample.
`parallel_attempts`	How many counterfactuals are created for the same factual in parallel.
`predictor_path`
`proximity_weight`
`sparsity_weight`
`y_target_goal_confidence`

explainer_type: str = 'DiCEExplainer'¶: The type of explanation that shall be used. This is necessary to know which pydantic class to use when loading from yaml.

num_attempts: int = 1¶: The number of counterfactuals created for the same sample.

parallel_attempts: int = 1¶: How many counterfactuals are created for the same factual in parallel.