shapiq
shapiq copied to clipboard
mmschlk
Shapley Interactions for Machine Learning
Add the correct paper references as links to the docstrings of the approximators and explainers (in case of treeshapiq).
Write an example notebook showcasing how to use TreeSHAP-IQ for SOTA tree models. An illustrative example is the California housing dataset which shows easy to understand interactions in the latitude...
Make plotting functions also callable from the explainer class with the last results from the explain function. Do this after #34.
### Description Add functionality to the package to allow for calculation of interactions for multiple instances/games to allow for the following scenario. > A user wants to put a 2d...
Add the interventional interaction method from this [paper](https://ojs.aaai.org/index.php/AAAI/article/view/26322). This would be another algorithm accessible through the `TreeExplainer`. The method should work with our `TreeModel` data structure (potentially extend `TreeModel`). This...
So that the user can define a game by a value function (using the model and data), which will then allow computing interaction values.
We need an easy-to-find guide on how to add custom tree models that we are not nativlely supporting to the repository. This should be accessible in the documentation. I think...
Run an extensive benchmark of all approximators and measure the approximation quality in terms of MSE (and other loss functions).
We should add mypy static type checking. This will however need some changes in what typehints we currently are using. Especially there will be problems/changes with current np.ndarry[int] kind of...