Despite ongoing theoretical research on cross-validation (CV), many theoretical questions about CV remain widely open. This motivates our investigation into how properties of algorithm-distribution pairs can affect the choice for the number of folds in k-fold cross-validation. Our results consist of a novel decomposition of the mean-squared error of cross-validation for risk estimation, which explicitly captures the correlations of error estimates across overlapping folds and includes a novel algorithmic stability notion, squared loss stability, that is considerably weaker than the typically required hypothesis stability in other comparable works. Furthermore, we prove: 1. For every learning algorithm that minimizes empirical error, a minimax lower bound on the mean-squared error of k-fold CV estimating the population risk LD: \ ₊ ₍\; ₃\; E\![ (L₂ₕ^ (k) - L₃) ^2 \;=\; Ω\! (k/n), \] where n is the sample size and k the number of folds. This shows that even under idealized conditions, for large values of k, CV cannot attain the optimum of order 1/n achievable by a validation set of size n, reflecting an inherent penalty caused by dependence between folds. 2. Complementing this, we exhibit learning rules for which \ ₃\; E\![ (L₂ₕ^ (k) - L₃) ^2 \;=\; Ω (k/n), \] matching (up to constants) the accuracy of a hold-out estimator of a single fold of size n/k. Together these results delineate the fundamental trade-off in resampling-based risk estimation: CV cannot fully exploit all n samples for unbiased risk evaluation, and its minimax performance is pinned between the k/n and k/n regimes.
Building similarity graph...
Analyzing shared references across papers
Loading...
Ido Nachum
Rüdiger Urbanke
Thomas Weinberger
Building similarity graph...
Analyzing shared references across papers
Loading...
Nachum et al. (Wed,) studied this question.
www.synapsesocial.com/papers/690fdcdaf60c54d04ea3821b — DOI: https://doi.org/10.48550/arxiv.2511.03554