This module contains the implementations of all valuation methods.
This includes semi-values, model-specific methods and any related algorithms. For several of the semi-value methods there are two implementations:
- A generic one combining some of the available samplers with a subclass of SemivalueValuation which defines a coefficient function.
- A dedicated class that chooses the right sampler to avoid numerical issues when computing otherwise necessary importance sampling corrections. You can read more about this in Sampling strategies for semi-values.
Info
For a full list of algorithms see: All data valuation methods implemented.