SHAPIQ Approximator

General Monte Carlo approximator for any-order interactions using SHAPIQ Muschalik et al.[1].

from __future__ import annotations

import numpy as np

import shapiq

N_PLAYERS = 8
BUDGET = 200
feature_names = [f"x{i}" for i in range(N_PLAYERS)]

weights = np.array([0.4, 0.3, 0.2, 0.1, 0.05, -0.1, -0.2, -0.3])


def game_fun(coalitions: np.ndarray) -> np.ndarray:
    coalitions = np.atleast_2d(coalitions)
    return (coalitions @ weights) + 0.5 * coalitions[:, 0] * coalitions[:, 1]

Approximate k-SII values

approximator = shapiq.SHAPIQ(n=N_PLAYERS, max_order=2, index="k-SII", random_state=42)
iv = approximator.approximate(BUDGET, game_fun)
print(iv)

InteractionValues(
    index=k-SII, max_order=2, min_order=0, estimated=True, estimation_budget=200,
    n_players=8, baseline_value=0.0,
    Top 10 interactions:
        (0,): 0.37035788690476207
        (0, 1): 0.36198048941798944
        (1,): 0.26262475198412694
        (2,): 0.19381919642857148
        (1, 7): 0.17404877645502653
        (0, 7): 0.10531671626984127
        (2, 3): 0.10196312830687834
        (3,): 0.09848883928571427
        (6,): -0.2172036210317462
        (7,): -0.24692981150793666
)

Force plot

iv.plot_force(feature_names=feature_names)

Network plot

iv.plot_network(feature_names=feature_names)

References

[1]

Maximilian Muschalik, Fabian Fumagalli, Barbara Hammer, and Eyke Hüllermeier. Beyond treeshap: efficient computation of any-order shapley interactions for tree ensembles. In Thirty-Eighth AAAI Conference on Artificial Intelligence, (AAAI 2024), 14388–14396. AAAI Press, 2024. doi:10.1609/AAAI.V38I13.29352.