Proof of Theorem 5.2, p. 18 — (5.12) divided by δₙ: f(0,i,Qⁿᵢ) + H·Qⁿᵢ ≥ f(0,i,Q^{u,n}ᵢ) + H·Q^{u,n}ᵢ
ProvedStrongWeakEq.Limit.display_5_12_generator_formcontinuous-time-markov-chainequilibrium-controlp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1time-discretizationtime-inconsistency
Fix a mesh , a payoff , a vector , a state and two matrices . Write and as in (5.10), and as in (5.9). Then
Applied with , this turns the discrete one-step inequality (5.12) into an inequality between generator rows, which is the form in which the proof of Theorem 5.2 lets ; the limit is the first-order condition (3.10).
Formalization Note The equivalence is stated for arbitrary real matrices and vectors : it is pure algebra (, so both sides differ by the common term and the factor ). States are Fin N.
Preamble
import Mathlib import Definitions.Def_StrongWeakEq_Existence_Model import Definitions.Def_StrongWeakEq_Discrete_DiscreteModel import Definitions.Def_StrongWeakEq_Limit_Discretization
Formal statement
namespace StrongWeakEq.Limit
/-- Proof of Theorem 5.2, p. 18: for a mesh `δ > 0` and any vector `H`, the one-step inequality
(5.12) between `u` and `uⁿ` is equivalent to its generator form
`f(0,i,Q^{uⁿ}ᵢ) + H·Q^{uⁿ}ᵢ ≥ f(0,i,Q^{u}ᵢ) + H·Q^{u}ᵢ` with `Q^{u} = (u − I)/δ`. -/
theorem display_5_12_generator_form {N : ℕ} (f : ℝ → Fin N → (Fin N → ℝ) → ℝ)
(δ : ℝ) (hδ : 0 < δ) (H : Fin N → ℝ) (u un : Matrix (Fin N) (Fin N) ℝ) (i : Fin N) :
(kappaN f δ 0 i (u i) + H ⬝ᵥ u i ≤ kappaN f δ 0 i (un i) + H ⬝ᵥ un i) ↔
(f 0 i (genOf δ u i) + H ⬝ᵥ genOf δ u i ≤ f 0 i (genOf δ un i) + H ⬝ᵥ genOf δ un i) := by sorry
end StrongWeakEq.Limit
Source
Huang & Zhou, Strong and Weak Equilibria for Time-Inconsistent Stochastic Control in Continuous Time, arXiv:1809.09243v3, p. 18, proof of Theorem 5.2
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.