Original title: Exact Risk-Complexity Laws for Projective Boundaries in Scenario Optimization and Distribution-Free Certification
Giuseppe C. Calafiore · September 2026
When computers make decisions or predictions from limited data, we want to know the risk of making an error on future observations. Past methods showed that this error risk often follows a precise mathematical pattern called a beta lawA mathematical pattern describing the probability distribution of risk in prediction sets., using fixed dimensions of the data. However, the number of key data points shaping the final decision can vary at random. The researchers found the underlying mechanism that creates these risk formulas. They defined rules where a boundary of key points fixes the allowed choices, meaning unused accepted data can be removed without changing the boundary. By studying how this boundary behaves across different samples, they showed how the risk works when boundary sizes change. A stable pattern gives the classic beta law, while changing patterns require an exact correction. They also showed why looking only at observed complexity is not enough.
Imagine setting a speed limit on a highway based on observing a few cars to gauge safe speeds. Normally, you'd use a fixed rule to estimate risk, but if the number of cars you observe changes unpredictably each day, that rule becomes unreliable. Here, the decision rule is like the speed limit, and the boundary points are the key cars that define it. When the boundary size varies randomly, the method adjusts the risk formula to keep safety guarantees accurate.
This affects people using prediction systems with safety guarantees, like in finance or healthcare. Understanding the exact risk helps set more reliable confidence levels when data variability matters.
The abstract doesn't list explicit limitations, but the framework assumes proper projective boundary schemes and may not apply to all distribution-free methods. It also notes that observed complexity alone is insufficient for certificates.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Exact Risk-Complexity Laws for Projective Boundaries in Scenario Optimization and Distribution-Free Certification — Giuseppe C. Calafiore
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.