Proof of Proposition 4.1, p. 426 — a ⊗ b is the unique solution of min_{P ∈ U(a,b)} −H(P)
ProvedCompOT.EntropicLimit.entropy_max_at_productentropyindependenceoptimal-transportp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1
Let and . The product coupling lies in and is the unique solution of
every other satisfies .
This identifies the limit of as in Proposition 4.1: the coupling of maximal entropy between two prescribed marginals is the law of two independent random variables.
Formalization Note uses ; indices are 0-based.
Preamble
import Mathlib import Definitions.Def_CompOT_EntropicLimit_Defs
Formal statement
namespace CompOT.EntropicLimit
/-- Proof of Proposition 4.1, p. 426: for `a ∈ Σ_n`, `b ∈ Σ_m`, the solution of
`min_{P ∈ U(a,b)} -H(P)` is `a ⊗ b = (a_i b_j)_{i,j}`: it lies in `U(a,b)` and every other
coupling has strictly smaller entropy. -/
theorem entropy_max_at_product {n m : ℕ}
(a : Fin n → ℝ) (b : Fin m → ℝ)
(ha : a ∈ stdSimplex ℝ (Fin n)) (hb : b ∈ stdSimplex ℝ (Fin m)) :
(Matrix.of fun i j => a i * b j) ∈ CompOT.Assignment.couplings a b ∧
∀ Q ∈ CompOT.Assignment.couplings a b, Q ≠ (Matrix.of fun i j => a i * b j) →
entropy Q < entropy (Matrix.of fun i j => a i * b j) := by sorry
end CompOT.EntropicLimit
Source
Peyré & Cuturi, Computational Optimal Transport (FnT ML 2019), proof of Proposition 4.1, p. 426 (last display)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.