Comments (6)
Thanks @glemaitre for pointing it out. The primary goal of this FastTreeSHAP
package is to develop a fast implementation of the TreeSHAP algorithm (feature_perturbation="tree_path_dependent") that reproduce the results from the original implementation of the TreeSHAP algorithm (feature_perturbation="tree_path_dependent") in the SHAP package. For those users who are currently implementing the TreeSHAP algorithm (feature_perturbation="tree_path_dependent") in SHAP package, we aim to provide them with a more efficient way of implementing it.
The potential issue you pointed out is relevant to the mathematical details of the TreeSHAP algorithm in the original paper, therefore we think it is more appropriate to discuss this issue with the original authors of TreeSHAP. If the original authors update the TreeSHAP algorithm (feature_perturbation="tree_path_dependent") in the SHAP package, we will update our implementations on our side to make sure they produce exactly the same results.
from fasttreeshap.
@glemaitre I think you meant "on correlated features".
from fasttreeshap.
Thanks, I corrected the title.
from fasttreeshap.
Actually: "TreeSHAP does not exact SHAP values" => "TreeSHAP does not compute exact SHAP values"
from fasttreeshap.
Just to give some feedback here. The issue with the tree_path_dependent
approach has been discussed in the literature, e.g. Janzing et al.
The only way to not break the symmetry axiom is to perturb features in a marginal manner and not a conditional one. So I assume that there is nothing that can be done here.
from fasttreeshap.
Yes that's correct.
from fasttreeshap.
Related Issues (20)
- Plotting example HOT 5
- Cannot build fasttreeshap in linux environment HOT 1
- How to build FastTreeSHAP Wheel on windows, using Python 3.10.7
- Numpy<1.22 requirement, could we upgrade it? HOT 3
- SHAP Values Change and Additivity Breaks on NumPy Upgrade HOT 1
- The additivity check failed with feature_perturbation=“tree_path_dependent” HOT 5
- Parallelism not working when model_output="logloss" HOT 3
- Beeswarm plot colorbar is too narrow on jupyter notebooks HOT 1
- The notebook example is not working HOT 2
- 'numpy' has no attribute 'bool' HOT 1
- NumbaDeprecationWarning HOT 5
- "error: could not find a version that satisfies the requirement setuptools (from versions: none)" HOT 1
- ERROR IN INSTALLING FastTreeSHAP HOT 2
- FastTreeSHAP summary_plot plots interaction value instead of impact on model output
- "shap" as a dependency?
- Additivity check fails with XGBoost
- Update package
- xgboost version compatible with FastTreeSHAP
- Support for RandomForestQuantileRegressor
- cannot slice shap values like SHAP package HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from fasttreeshap.