Decision-Focused Learning in Network Interdiction Games
2026-08-10 • Computer Science and Game Theory
Computer Science and Game TheoryMachine Learning
AI summaryⓘ
The authors examine a type of game where one player defends a network and another tries to find the shortest path to attack it. They find that the usual decision-focused learning method, which combines prediction and decision-making, can fail because multiple predictors look equally good but perform poorly in actual interdiction. To fix this, they propose Adversarial Decision-Focused Learning (A-DFL), which trains the model using scenarios that simulate attacks, improving real-world performance. Their tests on both made-up and real networks show that A-DFL successfully restores the benefits of decision-focused learning in this setting.
decision-focused learningshortest-path network interdictionStackelberg gamemachine learningadversarial trainingend-to-end optimizationnetwork securitygame theorypredictor equivalenceinterdiction scenario
Authors
Luca M. Hartmann, Parinaz Naghizadeh
Abstract
We study decision-focused learning (DFL) in shortest-path network interdiction (SPNI) games, a Stackelberg game where an interdictor (leader) strengthens the networks' arcs against attacks, while an evader (follower) who is uncertain about costs of attacking network arcs relies on a machine-learned predictor to identify the shortest path. While DFL is highly effective as an end-to-end optimization framework, we show that it faces a fundamental structural failure when employed in this game setting: its training objective admits a broad decision-equivalence class of cost estimators that achieve zero nominal loss yet fail under interdiction, reversing DFL's usual advantage over a naive prediction-focused learning (PFL) approach. To address this, we propose Adversarial DFL (A-DFL), which replaces nominal training samples with interdicted scenarios to collapse the harmful equivalence class. Experiments on synthetic and real-world networks confirm that A-DFL restores DFL's advantage in this game setting, enabling effective end-to-end optimization.