P-ademacher Mini Calculator
Toy calculator for residue-based complexity on finite samples. This page estimates two expectations over random labels: a supremum-style distance (as written in one thesis draft definition) and a minimum mismatch fit score.
Interpretation
With residue labels in {0,...,p-1}, p-adic distance of two distinct residues is usually 1 (and 0 when equal). So these toy statistics reduce to mismatch-style quantities on finite samples.
- E[sup_f avg |f(x_i)-sigma_i|_p]: higher means some hypothesis can stay far from random labels.
- E[inf_f mismatch_rate]: lower means class can fit random labels better.