Lemma 2, p. 10 — the mean estimate equals the true mean decision
ProvedAggGameNet.Sync.lemma_2aggregate-trackingconsensusdistributed-algorithmp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let . If every matrix has unit column sums, the estimates start at , and they obey , then for every ,
This identity lets the analysis replace the network-wide average estimate with the actual average decision. It needs no projection or graph hypothesis.
Preamble
import Mathlib import Definitions.Def_AggGameNet_Sync_Setting
Formal statement
namespace AggGameNet.Sync
/-- Lemma 2, p. 10: column-stochastic averaging preserves the actual mean. -/
theorem lemma_2 {N n : ℕ} (hN : 0 < N)
(W : ℕ → Matrix (Fin N) (Fin N) ℝ)
(x v : ℕ → Fin N → E n)
(hcol : ∀ k i, ∑ j, W k j i = 1)
(hinit : v 0 = x 0)
(hrec : ∀ k i, v (k + 1) i = vhat W v k i + x (k + 1) i - x k i) :
∀ k, yavg v k = (1 / (N : ℝ)) • ∑ i, x k i := by sorry
end AggGameNet.Sync
Source
Koshal, Nedić, Shanbhag, Distributed Algorithms for Aggregative Games on Graphs, arXiv:1605.00267v2, Lemma 2, p. 10
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.