Answers are grounded in this paper and cite their sources.
In the traditional approach to assessing weapon effectiveness, the analyst first proposes an impact point distribution and a damage function and then integrates them to produce a single-shot kill probability. This approach is appropriate when the shooting data on a weapon are nonexistent or exceedingly scarce. However, this approach has an inherent limitation: when the true impact point distribution does not align with the proposed impact point distribution, the predicted weapon effectiveness can be misleading, jeopardizing critical military missions. Based on conformal prediction, a general machine learning strategy for generating finite-sample valid prediction regions, this article proposes a method to address this issue when shooting data are available. The proposed method is non-parametric by nature and does not require the analyst to specify any impact point distribution. The new method enables an analyst to obtain a prediction kill zone whose coverage probability is no less than a pre-assigned confidence level.
No takes yet. Share an insight, caveat, or question.
Liang Hong (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: