Module eval 

The filter’s numbers, aggregate only: how much ham it would call spam and how much spam it would catch at your two thresholds, each with its 95 % exact binomial interval (Clopper–Pearson), the share it calls unsure, and the area under the ROC curve. Nothing here names a message: these numbers may be read by an AI agent, so no subject, sender or word ever enters them.

Structs

AtThreshold

What one threshold does: messages at or above it are called spam.

Numbers

The evaluation of probabilities against labels.

Share

count of of, with its 95 % interval.

Constants

STRICT

The share of ham a strict threshold may call spam: one in two hundred.

Functions

at

What one threshold does to (probability of spam, is spam) pairs.

auc

The area under the ROC curve: the chance a random spam scores above a random ham (ties count half), by ranks (Mann–Whitney).

beta_fraction 🔒

The continued fraction of the incomplete beta function (modified Lentz).

beta_quantile 🔒

The x where the Beta(a, b) distribution reaches p, by bisection.

clopper_pearson

The exact (Clopper–Pearson) interval of k successes in n trials.

evaluate

Evaluates (probability of spam, is spam) pairs at the two thresholds.

grid

The same at each of thresholds, so that a threshold can be chosen (sioul spam eval’s grid): what each would set aside, and catch.

incomplete_beta 🔒

The regularized incomplete beta function I_x(a, b), by its continued fraction (Numerical Recipes, betai), with the symmetry for large x.

ln_gamma 🔒

ln Γ(x) for x > 0 (Lanczos, g = 7, nine terms).

strict

The lowest threshold at which at most rate of the ham is called spam: just above the score of the ham that would be one too many. None without ham.