Privacy-preserving ROC/PR curve approximation for federated learning.

A framework for preprocessing tabular data in federated learning.

Order-preserving transforms between Euclidean Distance, Inner Product, and Cosine Similarity.

Numerically stable power transforms with an extension to federated learning.

Parametric methods that normalize data but require careful handling to avoid numerical instability.

Shape-preserving interpolation method that preserves monotonicity and avoids overshoots.

Graphical tools for evaluating classification models, highlighting trade-offs in model performance.

Mathematically equivalent formulas for variance can have very different numerical stability.