pyreal.RealApp.produce_similar_examples#

RealApp.produce_similar_examples(x_orig, model_id=None, x_train_orig=None, y_train=None, format_output=True, num_examples=3, standardize=False, fast=True, format_y=True, algorithm=None, force_refit=False)[source]#

Produce a SimilarExamples explainer

Parameters:

  • x_orig (DataFrame) – Input to explain

  • model_id (string or int) – ID of model to explain

  • x_train_orig (DataFrame) – Data to fit on, if not provided during initialization

  • y_train (DataFrame or Series) – Training targets to fit on, if not provided during initialization

  • format_output (Boolean) – No functionality, included for consistency

  • num_examples (int) – Number of similar examples to return

  • standardize (Boolean) – If True, standardize data before using it to get similar examples. Recommended if model-ready data is not already standardized

  • fast (Boolean) – If True, use a faster algorithm for generating similar examples. Disable if faiss is not available

  • format_y (Boolean) – If True, format the ground truth y values returned using self.pred_format_func

  • algorithm (string) – Name of algorithm

  • force_refit (Boolean) – If True, initialize and fit a new explainer even if the appropriate explainer already exists

Returns:

DataFrame, “y”: Series, “Input”: Series} (if series),

else {“id” -> {“X”: DataFrame, “y”: Series, “Input”: Series}}

X is the examples, ordered from top to bottom by similarity to input and y is the corresponding y values Input is the original input in the same feature space

Return type:

{“X”