Proposition C.2, p.25 — minima of a nondecreasing sequence of lsc functions on a compact set
ProvedRhoArbitrage.proposition_C_2arbitragemathematical-financep2o-gran-per-chapterp2o-plan-paperp2o-v1portfolio-optimizationrisk-measuresstar-shaped-risk
Let be compact and let be a nondecreasing sequence of lower semi-continuous functions with for all . Then
Furthermore, if is a sequence with , then any limit point of is a minimiser of .
This is the exchange of limit and minimum that identifies with the limit of .
Formalization Note Attainment of each minimum is stated explicitly.
Differences from the source:
- encoding: K ⊂ X compact, X a topological space, f_t : K → [−∞, ∞] → a type
Kwith[TopologicalSpace K] [CompactSpace K] [Nonempty K]; reason: the functions live on K, so only the subspace topology matters; every compact subset of a space is such a type, and "min" presupposes K ≠ ∅ - encoding: min_{x∈K} in the display → explicit attainment (
IsMinOn) together with⨅ x, _; reason: the printed min asserts attainment, and then it equals the infimum - encoding: limit point of (x_t) →
MapClusterPt y atTop xs; reason: a limit point of a sequence is a cluster point - encoding: indices t ≥ 1 → indices
t : ℕ; reason: a shift of the index
Preamble
import Mathlib open Filter Topology
Formal statement
namespace RhoArbitrage
/-- **Proposition C.2 (minima of an increasing sequence of lsc functions).** Herdegen–Khan,
arXiv:2202.07610v3, Proposition C.2, p.25.
Let `K` be compact and `f_t : K → [-∞, ∞]` a nondecreasing sequence of lower semi-continuous
functions with pointwise limit `F`. Then `min_K F = lim_t min_K f_t`; and if `(x_t)` is a sequence
with `min_K f_t = f_t(x_t)`, then any limit point of `(x_t)` is a minimiser of `F`.
**Formalization Note.** The functions live on `K` as printed, so `K` is a type carrying the
compact topology; a compact subset `K ⊂ X` of a topological space, with its subspace topology, is
an instance. "`min`" presupposes `K ≠ ∅`, hence `[Nonempty K]`. The first two conjuncts are the
attainment of the minima that the printed `min` asserts, and each `min` in the display is then the
infimum `⨅ x, _`. A limit point is a cluster point (`MapClusterPt`). The paper indexes by `t ≥ 1`,
Lean by `t : ℕ`. -/
theorem proposition_C_2 {K : Type*} [TopologicalSpace K] [CompactSpace K] [Nonempty K]
(f : ℕ → K → EReal) (F : K → EReal) (hlsc : ∀ t, LowerSemicontinuous (f t))
(hmono : ∀ x, Monotone (fun t => f t x))
(hlim : ∀ x, Tendsto (fun t => f t x) atTop (𝓝 (F x))) :
(∀ t, ∃ x, IsMinOn (f t) Set.univ x) ∧
(∃ x, IsMinOn F Set.univ x) ∧
Tendsto (fun t => ⨅ x, f t x) atTop (𝓝 (⨅ x, F x)) ∧
∀ xs : ℕ → K, (∀ t, IsMinOn (f t) Set.univ (xs t)) →
∀ y, MapClusterPt y atTop xs → IsMinOn F Set.univ y := by sorry
end RhoArbitrage
Source
Herdegen, Khan, ρ-arbitrage and ρ-consistent pricing for star-shaped risk measures, arXiv:2202.07610v3, Proposition C.2, p.25
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.