You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[FORM-S13] Optimal pooling and Dobrushin consensus for collective inference (fep-110..113) #96
Consensus uses one fixed 2-agent mixing matrix and fep-110's pooled mass is rfl. Prove the log-pool and linear pool are the unique minimisers of the two weighted-KL objectives (via the Gibbs principle), and Dobrushin contraction for any scrambling row-stochastic W.
Acceptance
A 3-agent asymmetric W witness and a non-scrambling countermodel; the fixed 2-agent kernel is an instance.
Failure controls and claim boundary
No disposition promotion from compilation, import or numerical agreement; the maturity record changes only after independent semantic review. No duplicate topics and no automatic sealed-roster expansion.
Closure requires actual exact-source commands, results and retained receipts, with independent review where specified. Remove its open TODO row only after acceptance; retain delivery facts in changelog/history. Size describes implementation scope, not a semantic-version promise.
Status update (2026-10-09). Delivered in #162 (via #164), merged to main: Linear-pool and log-pool KL decompositions with unique-minimiser theorems, Dobrushin contraction spread(Wx) ≤ (1−δ)·spread(x) for any finite row-stochastic W, the existing 2-agent kernel as an instance, an asymmetric 3-agent strict-contraction witness, and an identity countermodel.
Evidence: lake --wfail build FepSketches passes at the pin (Lean v4.34.1, Mathlib d13f23b7), every new declaration depends only on propext/Classical.choice/Quot.sound, and all generator --checks plus hosted CI pass.
Remaining for closure: independent semantic review of the delivered statements against the topic's theorem proxy. Per this issue's claim boundary, no disposition in config/theorem_maturity.yaml was changed, and none may change without that review.
Accepted on main via #171. In FEP.CollectiveInference: linearPool_kl_unique_minimiser and logPool_kl_unique_minimiser; valueSpread_mixStep_le for any row-stochastic W; the 3-agent asymmetric witness asymMatrix3_strict_contraction; the non-scrambling countermodel identityMixing2_no_contraction; and the 2-agent kernel as an instance (consensus_spread_le_half). Any maturity change still needs independent review. Row retired.
Activity
docxology commentedon Oct 9, 2026
Status update (2026-10-09). Delivered in #162 (via #164), merged to
main: Linear-pool and log-pool KL decompositions with unique-minimiser theorems, Dobrushin contractionspread(Wx) ≤ (1−δ)·spread(x)for any finite row-stochastic W, the existing 2-agent kernel as an instance, an asymmetric 3-agent strict-contraction witness, and an identity countermodel.Evidence:
lake --wfail build FepSketchespasses at the pin (Lean v4.34.1, Mathlibd13f23b7), every new declaration depends only onpropext/Classical.choice/Quot.sound, and all generator--checks plus hosted CI pass.Remaining for closure: independent semantic review of the delivered statements against the topic's theorem proxy. Per this issue's claim boundary, no disposition in
config/theorem_maturity.yamlwas changed, and none may change without that review.docxology commentedon Oct 11, 2026
Accepted on
mainvia #171. InFEP.CollectiveInference:linearPool_kl_unique_minimiserandlogPool_kl_unique_minimiser;valueSpread_mixStep_lefor any row-stochastic W; the 3-agent asymmetric witnessasymMatrix3_strict_contraction; the non-scrambling countermodelidentityMixing2_no_contraction; and the 2-agent kernel as an instance (consensus_spread_le_half). Any maturity change still needs independent review. Row retired.