Exact and empirical estimation of misclassification probability
This work addresses risk estimation for classification, but appears incremental as it focuses on analytic approximations for a specific classifier type.
The paper tackles the problem of risk estimation in classification by deriving analytic approximations for the maximum bias of empirical risk for a histogram classifier, and studies their use in empirical risk estimation.
We discuss the problem of risk estimation in the classification problem, with specific focus on finding distributions that maximize the confidence intervals of risk estimation. We derived simple analytic approximations for the maximum bias of empirical risk for histogram classifier. We carry out a detailed study on using these analytic estimates for empirical estimation of risk.