Work

A conditional Lean reduction for two-class Bernoulli mean exceedance

Year: 2026
Type: preprint
Abstract: This candidate manuscript presents a Lean reduction for mean exceedance in a weighted Bernoulli sum with two weight classes. The same capacity-minimizing witness selects a rational endpoint and gives containment of the completed winning event in the original winning event. A fixed-event propagation inequality then transfers a bound at the selected bias to the original bias. For eight equal exceptional weights and n equal ordinary weights, the original-weight result is conditional on a retained finite premise. An exact checker establishes ordered key coverage and checks the strict integer inequality at each certified endpoint. Five certified multiplicities cover 29,841 of the 480,600 retained endpoints. Separate clean replays and a joint finite audit cover n = 80 and n = 99, accounting for 23,417 endpoints.
Language: English
Open Access status: green
OpenAlex ID: W7234253317
Cites: 2
Cited by:
Related to: 10