Exercise 8.3 — the -fair objective is strictly concave
ProvedKellyStochasticNetworks.alpha_fair_concaveFor every , positive weights and positive flow counts , the weighted -fair objective
is strictly concave on the open positive orthant.
Three regimes have to be checked and they look different: for the exponent is positive and the prefactor is positive, so each summand is a concave power; for both the exponent and the prefactor change sign, and the summand is again concave; and at the summand is a logarithm. Strict concavity holds throughout, which is what makes the -fair allocation unique.
Together with compactness of the feasible region this gives existence and uniqueness of the allocation, so that "the -fair rate allocation" names a single vector — and, since the problem generalizes of section 7.1, it is the same fact that made the proportionally fair allocation well defined there.
Formalization Note Strict concavity is asserted on the open positive orthant, a convex set, which is where real powers and the logarithm are meaningful. The book restricts attention to the coordinates with ; here every is assumed positive, which is the same restriction applied to the whole index set.
import Mathlib import Definitions.Def_KellyStochasticNetworks_Wardrop import Definitions.Def_KellyStochasticNetworks_Congestion import Definitions.Def_KellyStochasticNetworks_FlowLevel
namespace KellyStochasticNetworks
theorem alpha_fair_concave {R : ℕ} (w n : Fin R → ℝ) (α : ℝ) (hα : 0 < α)
(hw : ∀ r, 0 < w r) (hn : ∀ r, 0 < n r) :
StrictConcaveOn ℝ {X : Fin R → ℝ | ∀ r, 0 < X r} (alphaFairObjective w n α) := by sorry
end KellyStochasticNetworksConfirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.