Two-Feature Shapley Values

Problem

Compute exact Shapley values for a two-feature model with an interaction. The empty-set value is 0, the individual feature values are v1 and v2, and the joint value is v12. Feature 1 gets 0.5*v1 + 0.5*(v12 - v2) and feature 2 gets 0.5*v2 + 0.5*(v12 - v1). Return [shap1, shap2].

Hint

With only two features each Shapley value averages the marginal contribution over both orderings.

Test cases

case 1

in: [1,2,4]

expected: [1.5,2.5]

case 2

in: [0,0,0]

expected: [0,0]

case 3

in: [3,1,3]

expected: [2.5,0.5]

case 4

in: [-1,2,1]

expected: [-1,2]

Starter codePython

def shapley_two_features(v1, v2, v12):
    # Returns [shap1, shap2]
    # Your code here
    pass

4 test cases · run your code against them in the editor