An FPRAS for Cell-Bounded Contingency Tables

OpenAI

One randomized algorithm counts nonnegative integer matrices with prescribed margins and individual cell bounds, including zeros, to relative error ε\varepsilon with failure probability at most δ\delta . Both dimensions and all numeric inputs vary in binary, and runtime is polynomial in input length, ε−1\varepsilon ^{-1} and log⁡δ−1\log \delta ^{-1} on every run.

What was already known

Prior approximate counting covers fixed row count or restricted margins. The cataloged exact sampler has no individual cell bounds and is a distinct companion.

AI involvement AI-led

The manuscript credits OpenAI as author, and the pinned release attributes its manuscripts to the internal model. Individual session details are undocumented.

Importance Notable

A fully polynomial approximate counter for varying dimensions with arbitrary binary-encoded bounds.

Technical verification Formal (Lean)

OpenAI’s selected Lean scope explicitly includes the cell-bounded FPRAS and zero-count guarantee in addition to the separate sampling theorem.

Formal checking

OpenAI’s selected Lean scope explicitly includes the cell-bounded FPRAS and zero-count guarantee in addition to the separate sampling theorem.

Checker
OpenAI’s reported Lean formalization
Scope
OpenAI’s selected Lean scope explicitly includes the cell-bounded FPRAS and zero-count guarantee in addition to the separate sampling theorem.
Basis
Reported by the linked checker

Human understanding Few

The reviewed sources do not document a human explanation of the assessed contribution.

Correction and publication history