responsibility¶
calculate causal responsibility
Functions¶
|
Split a box into 4 contiguous children or None if no possible. |
|
Update the max tree depth reached. |
|
Reduce the search queue to improve both efficiency and (possibly) result quality. |
|
Calculate causal responsiblity. |
Module Contents¶
- responsibility.subbox(tree, name, max_depth, min_size, mode, r_map=None)¶
Split a box into 4 contiguous children or None if no possible.
@param tree: a search tree @param name: a node name to search for in <tree> @param max_depth: tree depth limit for new children @param min_size: minimum new child size @param mode: spectral, tabular, RGB, L or voxel @param r_map=None: responsibility map
@return None or new children
- responsibility.update_depth_reached(depth_reached, passing)¶
Update the max tree depth reached.
@param depth_reached: current max_depth @param passing: a list of passing Mutant
@return int
- responsibility.prune(mutants, technique=Queue.Intersection, keep=None)¶
Reduce the search queue to improve both efficiency and (possibly) result quality.
@param mutants: a list of passing Mutant objects @param technique: a Queue enum @param keep=None: how many items to keep in the queue, all if keep is None
@return a list of mutants of length <= keep
- Parameters:
mutants (List[rex_xai.mutants.mutant.Mutant])
- responsibility.causal_explanation(process, data, args, prediction_func, current_map=None)¶
Calculate causal responsiblity.
@param process: an integer value @param data: a Data object @param args: a CausalArgs object @param prediction_func: a higher order
function that calls a model and return a Prediction object
- Parameters:
data (rex_xai.input.input_data.Data)
args (rex_xai.input.config.CausalArgs)
