SHAP: Linear Model Attributions
Beginner
The problem, from first principles
For a linear model with independent features, the Shapley value of a feature is its weight times how far its value is from the average. This gives a closed form that matches the exact computation, a useful check on the general method.
From theory to code
Implement linear_shap(w, x, mean), returning w * (x - mean) element-wise.
Constraints
meanis the average of each feature over background data.
Hints
Hint 1
Subtract the means from the input, then multiply by the weights.
Keywords
research-papersclassical-mlinterpretabilityshap
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.