for
ProvedRankOneBTF.binomial_reciprocal_meanexact-recoverylinear-programmingp2o-gran-per-chapterp2o-plan-paperp2o-v1rank-one-btftensor-factorization
Let be a binomial random variable with parameters , . Then
The identity evaluates the expected reciprocal of the uncorrupted counts in the proofs of Claims 3 and 7 (with ).
Formalization Note The expectation is written as the explicit binomial sum.
Differences from the source:
- encoding: a binomial parameter p̄ (the paper applies it with p̄ = 1 − p ∈ (0,1]) → 0 < p̄ ≤ 1; reason: at p̄ = 0 the right-hand side is undefined; this excludes only that case
- encoding: E[1/(1+X)] for X ~ Bin(N, p̄) → the explicit sum Σ_k C(N,k) p̄^k (1−p̄)^{N−k}/(1+k); reason: the definition of the expectation
Preamble
import Mathlib
Formal statement
namespace RankOneBTF
/-- **Binomial reciprocal identity** (arXiv:2202.07053v3, p.22, proof of Claim 3: "for a
binomial random variable X with parameters (N, p̄), we have
E[1/(1+X)] = (1/((N+1)p̄))(1 − (1 − p̄)^{N+1})"; also used in Steps 1 and 2 of Claim 7, p.36).
**Formalization Note** `E[1/(1+X)]` for `X ~ Bin(N, p̄)` is written as the explicit sum
`Σ_{k=0}^{N} C(N,k) p̄^k (1−p̄)^{N−k} / (1+k)`. The success probability satisfies
`0 < p̄ ≤ 1`: at `p̄ = 0` the right-hand side is undefined. -/
theorem binomial_reciprocal_mean (N : ℕ) (q : ℝ) (hq0 : 0 < q) (hq1 : q ≤ 1) :
∑ k ∈ Finset.range (N + 1),
(N.choose k : ℝ) * q ^ k * (1 - q) ^ (N - k) / (1 + (k : ℝ)) =
1 / (((N : ℝ) + 1) * q) * (1 - (1 - q) ^ (N + 1)) := by sorry
end RankOneBTF
Source
Del Pia, Khajavirad, Rank-one Boolean tensor factorization and the multilinear polytope, arXiv:2202.07053v3, p.22, proof of Claim 3: "for a binomial random variable X with parameters (N, p̄), we have E[1/(1+X)] = (1/((N+1)p̄))(1 − (1 − p̄)^{N+1})"; also Step 1 of Claim 7, p.36
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.