FA-73059 / Probabilistic sketches / Member archive

Count-Min sketch with conservative update: error bound omits the e factor · case 04

Callers trust estimates to be within a bound that Count-Min does not guarantee.

Member previewVariant 4 · 3 implementations · 6 checks per implementation

Case contract

Input {w, d, ops}. Row r hashes key to ((A_r*key + B_r) mod 2^31-1) mod w with fixed A and B. add with count <= 0 is "rejected" (conservative update cannot represent deletions). Otherwise the current estimate e is the minimum of the key's cells, and each cell is raised to max(cell, e + count). The stream total counts accepted mass and the overestimate bound is ceil(e/w * total). q returns the minimum of the key's cells. Return [results, total, bound, table].

Why this case matters

Conservative update reduces Count-Min overestimation in heavy-hitter monitors, but only if the raise-to-target rule is applied exactly.

One recorded failure

Sample boundary fixture

This sample comes from the broken implementation of a controlled reproducer.

Boundary fixture	Actual	Expected	Outcome
colliding keys	`[["ok", "ok", "ok", 5, 2, 0], 7, 2, [[5, 0, 0, 0, 2], [0, 2, 0, 0, 5], [0, 2, 0, 0, 5]]]`	`[["ok", "ok", "ok", 5, 2, 0], 7, 4, [[5, 0, 0, 0, 2], [0, 2, 0, 0, 5], [0, 2, 0, 0, 5]]]`	Failed

MEMBER ARCHIVE

The complete case is available to members.

This record includes three runnable implementations, regression fixtures, execution results, and source hashes.

Member access is invitation-based. Sign in with your invited account to inspect the sources.