🤖 AI Summary
This work addresses the interpretability of tree ensemble models by generating *minimal-cost exact explanations*. To overcome limitations of prior approaches—which enumerate all minimal explanations at prohibitive computational cost—the authors: (1) design an efficient formal verification oracle, accelerating verification by several orders of magnitude over prior work; (2) adapt the MARCO algorithm into m-MARCO, the first method capable of computing a single minimal-cost explanation (rather than enumerating all minimal ones); and (3) integrate SAT/MaxSAT modeling, minimal hitting set characterization, and incremental search optimization. Experiments demonstrate end-to-end speedups of up to 2×, while the generated minimal-cost explanations are substantially more concise—often constituting only a tiny fraction of the full set of minimal explanations—thereby significantly enhancing practicality and deployability.
📝 Abstract
The ability to explain why a machine learning model arrives at a particular prediction is crucial when used as decision support by human operators of critical systems. The provided explanations must be provably correct, and preferably without redundant information, called minimal explanations. In this paper, we aim at finding explanations for predictions made by tree ensembles that are not only minimal, but also minimum with respect to a cost function. To this end, we first present a highly efficient oracle that can determine the correctness of explanations, surpassing the runtime performance of current state-of-the-art alternatives by several orders of magnitude when computing minimal explanations. Secondly, we adapt an algorithm called MARCO from related works (calling it m-MARCO) for the purpose of computing a single minimum explanation per prediction, and demonstrate an overall speedup factor of two compared to the MARCO algorithm which enumerates all minimal explanations. Finally, we study the obtained explanations from a range of use cases, leading to further insights of their characteristics. In particular, we observe that in several cases, there are more than 100,000 minimal explanations to choose from for a single prediction. In these cases, we see that only a small portion of the minimal explanations are also minimum, and that the minimum explanations are significantly less verbose, hence motivating the aim of this work.