Documentation

Tdaf.Analysis.Convex.Convergence

Equi-Lipschitz families and convergence of convex functions #

Four theorems about families of convex functions on a relatively open convex set. A pointwise bounded family is uniformly bounded and equi-Lipschitzian on compact subsets; a function convex in x and continuous in t is jointly continuous; pointwise convergence on a dense subset propagates and becomes uniform on compact subsets; and a bounded sequence has a subsequence converging uniformly on compact subsets.

Each theorem appears twice: an interior form, on an open convex set, which carries the whole argument, and a _relint form, on ri C for an arbitrary convex C, obtained from it through the linear chart of Continuity.lean. The hypotheses are the weakened pair throughout: a subset C' with ri C ⊆ cl C' on which the family is pointwise bounded above, plus a single point of ri C at which it is bounded below. Taking C' = ri C recovers the headline statements, and the two convergence theorems need the weakened form, their own hypotheses being about a dense subset.

Main results #

Implementation notes #

The functions are real-valued rather than EReal-valued: these theorems are about families finite on a relatively open convex set, and every conclusion — a supremum, a Lipschitz constant, a uniform bound, a limit — is a statement about real numbers. So the family is f : ι → E → ℝ with ∀ i, ConvexOn ℝ C (f i), which composes directly with Mathlib; a caller holding an EReal-valued ConvexFn converts with ConvexFn.convexOn_toReal_dom. The upper-bound hypothesis actually needs only C ⊆ conv (cl C'), which is what bddAbove_range_of_subset_convexHull_closure proves; the theorems are stated with cl C' because the step from a bound to uniform convergence needs points of C' metrically near S, which a convex hull does not supply. The subsequence theorem avoids a diagonal argument: the values on a countable dense subset live in a compact box in ℕ → ℝ, compact by Tychonoff and sequentially compact because ℕ → ℝ is first countable.

References #

theorem Tdaf.ConvexAnalysis.convexOn_ciSup {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] {ι : Type u_2} {U : Set W} {f : ι → W → ℝ} [Nonempty ι] (hUc : Convex ℝ U) (hf : ∀ (i : ι), ConvexOn ℝ U (f i)) (hbdd : ∀ x ∈ U, BddAbove (Set.range fun (i : ι) => f i x)) :
ConvexOn ℝ U fun (x : W) => ⨆ (i : ι), f i x

The pointwise supremum of a family of functions convex on U is convex on U, provided the supremum is finite at every point of U.

theorem Tdaf.ConvexAnalysis.bddAbove_range_of_subset_convexHull_closure {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {ι : Type u_2} {U : Set W} {f : ι → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ι), ConvexOn ℝ U (f i)) {C' : Set W} (hC' : C' ⊆ U) (hdense : U ⊆ (convexHull ℝ) (closure C')) (hbdd : ∀ x ∈ C', BddAbove (Set.range fun (i : ι) => f i x)) (x : W) :
x ∈ U → BddAbove (Set.range fun (i : ι) => f i x)

Pointwise boundedness spreads through a convex hull. If a family of functions convex on an open convex set U is pointwise bounded above on a subset C' of U with U ⊆ conv (cl C'), then it is pointwise bounded above on all of U. This is the weakest form of the upper-bound hypothesis: the set of points where the family is bounded above is convex, so its closure contains conv (cl C'), and a convex set contains the interior of its own closure.

theorem Tdaf.ConvexAnalysis.exists_forall_le_of_isCompact {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {ι : Type u_2} {U : Set W} {f : ι → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ι), ConvexOn ℝ U (f i)) (hbdd : ∀ x ∈ U, BddAbove (Set.range fun (i : ι) => f i x)) {S : Set W} (hS : IsCompact S) (hSU : S ⊆ U) :
∃ (M : ℝ), ∀ (i : ι), ∀ x ∈ S, f i x ≤ M

Uniform boundedness from above: a family of functions convex on an open convex set U and pointwise bounded above there is uniformly bounded above on every compact subset of U. The pointwise supremum is a finite convex function, hence continuous, hence bounded on compact sets.

theorem Tdaf.ConvexAnalysis.exists_forall_ge_of_isBounded {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {ι : Type u_2} {U : Set W} {f : ι → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ι), ConvexOn ℝ U (f i)) (hbdd : ∀ x ∈ U, BddAbove (Set.range fun (i : ι) => f i x)) {x₀ : W} (hx₀ : x₀ ∈ U) (hbelow : BddBelow (Set.range fun (i : ι) => f i x₀)) {S : Set W} (hSb : Bornology.IsBounded S) (hSU : S ⊆ U) :
∃ (m : ℝ), ∀ (i : ι), ∀ x ∈ S, m ≤ f i x

Uniform boundedness from below: if a family of functions convex on an open convex set U is pointwise bounded above on U and bounded below at a single point x₀, it is uniformly bounded below on every bounded subset of U. For x ∈ U the point z = x₀ + (δ/‖x₀ - x‖) • (x₀ - x) lies on the sphere of radius δ about x₀, and x₀ is a convex combination of z and x, which gives a bound depending on x only through ‖x₀ - x‖.

Pointwise boundedness in the interior form #

theorem Tdaf.ConvexAnalysis.exists_forall_abs_le_of_isCompact {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {ι : Type u_2} {U : Set W} {f : ι → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ι), ConvexOn ℝ U (f i)) {C' : Set W} (hC' : C' ⊆ U) (hdense : U ⊆ closure C') (hab : ∀ x ∈ C', BddAbove (Set.range fun (i : ι) => f i x)) (hbe : ∃ x₀ ∈ U, BddBelow (Set.range fun (i : ι) => f i x₀)) {S : Set W} (hS : IsCompact S) (hSU : S ⊆ U) :
∃ (M : ℝ), 0 ≤ M ∧ ∀ (i : ι), ∀ x ∈ S, |f i x| ≤ M

The uniform boundedness half, in the interior form: a family of functions convex on an open convex set U, pointwise bounded above on a subset C' whose closure contains U and bounded below at a single point of U, is uniformly bounded on every compact subset of U. Taking C' = U recovers the headline statement.

theorem Tdaf.ConvexAnalysis.exists_forall_lipschitzOnWith_of_isCompact {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {ι : Type u_2} {U : Set W} {f : ι → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ι), ConvexOn ℝ U (f i)) {C' : Set W} (hC' : C' ⊆ U) (hdense : U ⊆ closure C') (hab : ∀ x ∈ C', BddAbove (Set.range fun (i : ι) => f i x)) (hbe : ∃ x₀ ∈ U, BddBelow (Set.range fun (i : ι) => f i x₀)) {S : Set W} (hS : IsCompact S) (hSU : S ⊆ U) :
∃ (K : NNReal), ∀ (i : ι), LipschitzOnWith K (f i) S

The equi-Lipschitz half, in the interior form: under the hypotheses of exists_forall_abs_le_of_isCompact a single Lipschitz constant works for every member of the family on every compact subset of U. One ε-collar and one bound M feed ConvexOn.lipschitzOnWith_of_abs_le_of_cthickening_subset, whose constant 2M/ε does not mention the function.

Convergence from a dense subset #

theorem Tdaf.ConvexAnalysis.uniformCauchySeqOn_of_dense {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {U : Set W} {f : ℕ → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ℕ), ConvexOn ℝ U (f i)) {C' : Set W} (hC' : C' ⊆ U) (hdense : U ⊆ closure C') (hab : ∀ x ∈ C', BddAbove (Set.range fun (i : ℕ) => f i x)) (hbe : ∃ x₀ ∈ U, BddBelow (Set.range fun (i : ℕ) => f i x₀)) (hcau : ∀ x ∈ C', CauchySeq fun (i : ℕ) => f i x) {S : Set W} (hS : IsCompact S) (hSU : S ⊆ U) :

The uniform Cauchy property. A sequence of functions convex on an open convex set U which is pointwise Cauchy on a subset C' whose closure contains U is uniformly Cauchy on every compact subset of U. Given ε, equi-Lipschitz continuity supplies one Lipschitz constant for the whole sequence on a compact collar of S, and finitely many points of C' then suffice.

theorem Tdaf.ConvexAnalysis.exists_tendstoUniformlyOn_of_dense {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {U : Set W} {f : ℕ → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ℕ), ConvexOn ℝ U (f i)) {C' : Set W} (hC' : C' ⊆ U) (hdense : U ⊆ closure C') (hcv : ∀ x ∈ C', ∃ (L : ℝ), Filter.Tendsto (fun (i : ℕ) => f i x) Filter.atTop (nhds L)) :
∃ (g : W → ℝ), ConvexOn ℝ U g ∧ (∀ x ∈ U, Filter.Tendsto (fun (i : ℕ) => f i x) Filter.atTop (nhds (g x))) ∧ ∀ ⦃S : Set W⦄, IsCompact S → S ⊆ U → TendstoUniformlyOn f g Filter.atTop S

Convergence from a dense subset, in the interior form: a sequence of functions convex on an open convex set U which converges pointwise on a subset C' whose closure contains U converges pointwise on all of U, the limit is convex, and the convergence is uniform on every compact subset of U.

Taking C' = U gives the version in which convergence is assumed everywhere.

theorem Tdaf.ConvexAnalysis.tendstoUniformlyOn_of_tendsto {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {U : Set W} {f : ℕ → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ℕ), ConvexOn ℝ U (f i)) {g : W → ℝ} (hg : ∀ x ∈ U, Filter.Tendsto (fun (i : ℕ) => f i x) Filter.atTop (nhds (g x))) {S : Set W} (hS : IsCompact S) (hSU : S ⊆ U) :

The same with the limit function supplied: pointwise convergence on all of an open convex U upgrades to uniform convergence on compact subsets.

theorem Tdaf.ConvexAnalysis.eventually_forall_le_add_of_eventually_le {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {U : Set W} {f : ℕ → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ℕ), ConvexOn ℝ U (f i)) {g : W → ℝ} (hg : ConvexOn ℝ U g) (hle : ∀ x ∈ U, ∀ δ > 0, ∀ᶠ (i : ℕ) in Filter.atTop, f i x ≤ g x + δ) {S : Set W} (hS : IsCompact S) (hSU : S ⊆ U) {ε : ℝ} (hε : 0 < ε) :
∀ᶠ (i : ℕ) in Filter.atTop, ∀ x ∈ S, f i x ≤ g x + ε

An eventual upper bound, in the interior form: if a sequence of functions convex on an open convex set U satisfies limsup_i f i x ≤ g x pointwise for a convex g, then on each compact S ⊆ U the bound f i ≤ g + ε holds for all large i. The limsup hypothesis is spelled as "for every ε > 0, eventually f i x ≤ g x + ε", which avoids the junk values Filter.limsup takes on unbounded sequences.

theorem Tdaf.ConvexAnalysis.exists_subseq_tendstoUniformlyOn {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {U : Set W} {f : ℕ → W → ℝ} (hU : IsOpen U) (hUc : Convex ℝ U) (hf : ∀ (i : ℕ), ConvexOn ℝ U (f i)) {C' : Set W} (hC' : C' ⊆ U) (hdense : U ⊆ closure C') (hbdd : ∀ x ∈ C', Bornology.IsBounded (Set.range fun (i : ℕ) => f i x)) :
∃ (φ : ℕ → ℕ) (g : W → ℝ), StrictMono φ ∧ ConvexOn ℝ U g ∧ (∀ x ∈ U, Filter.Tendsto (fun (i : ℕ) => f (φ i) x) Filter.atTop (nhds (g x))) ∧ ∀ ⦃S : Set W⦄, IsCompact S → S ⊆ U → TendstoUniformlyOn (fun (i : ℕ) => f (φ i)) g Filter.atTop S

Arzelà–Ascoli for convex functions, in the interior form: a sequence of functions convex on an open convex set U whose values are bounded at each point of a subset C' with U ⊆ cl C' has a subsequence converging, uniformly on every compact subset of U, to a finite convex function.

Joint continuity #

theorem Tdaf.ConvexAnalysis.continuousOn_prod_of_convexOn {W : Type u_1} [NormedAddCommGroup W] [NormedSpace ℝ W] [FiniteDimensional ℝ W] {U : Set W} {T : Type u_2} [TopologicalSpace T] [LocallyCompactSpace T] (hU : IsOpen U) (hUc : Convex ℝ U) {F : W × T → ℝ} (hconv : ∀ (t : T), ConvexOn ℝ U fun (x : W) => F (x, t)) {C' : Set W} (hC' : C' ⊆ U) (hdense : U ⊆ closure C') (hcont : ∀ x ∈ C', Continuous fun (t : T) => F (x, t)) :

Joint continuity, in the interior form: a real-valued function on U × T, with U open and convex and T locally compact, that is convex in its first argument and continuous in its second is jointly continuous. Continuity in t is only needed at the points of a subset C' of U whose closure contains U; taking C' = U gives the headline statement.

The chart: from interior to ri #

Every statement above is transported to the relative interior by the linear chart of Continuity.lean: exists_chart_retraction produces a subspace V, a continuous linear retraction r : E →L[ℝ] V, and the identity ri C = x₀ + ι (int (chart C x₀ V)). The three lemmas here are the bookkeeping that identity buys.

theorem Tdaf.ConvexAnalysis.convexOn_chart {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] {C : Set E} {x₀ : E} {V : Submodule ℝ E} {ψ : E → ℝ} (hψ : ConvexOn ℝ C ψ) :
ConvexOn ℝ (chart C x₀ V) fun (z : ↥V) => ψ (x₀ + ↑z)

A function convex on C is convex on the chart of C at x₀.

theorem Tdaf.ConvexAnalysis.mem_relint_of_mem_interior_chart {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] {C : Set E} {x₀ : E} {V : Submodule ℝ E} (himg : intrinsicInterior ℝ C = x₀ +ᵥ ⇑V.subtype '' interior (chart C x₀ V)) {z : ↥V} (hz : z ∈ interior (chart C x₀ V)) :

Points of the interior of the chart come from points of ri C.

theorem Tdaf.ConvexAnalysis.mem_interior_chart_of_mem_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] {C : Set E} {x₀ : E} {V : Submodule ℝ E} (himg : intrinsicInterior ℝ C = x₀ +ᵥ ⇑V.subtype '' interior (chart C x₀ V)) {x : E} (hx : x ∈ intrinsicInterior ℝ C) :
∃ z ∈ interior (chart C x₀ V), x₀ + ↑z = x

Points of ri C come from points of the interior of the chart.

theorem Tdaf.ConvexAnalysis.chart_subset_interior_chart {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] {C C' : Set E} {x₀ : E} {V : Submodule ℝ E} (himg : intrinsicInterior ℝ C = x₀ +ᵥ ⇑V.subtype '' interior (chart C x₀ V)) (hC' : C' ⊆ intrinsicInterior ℝ C) :
chart C' x₀ V ⊆ interior (chart C x₀ V)

A subset of ri C charts inside the interior of the chart.

theorem Tdaf.ConvexAnalysis.interior_chart_subset_closure_chart {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] {C C' : Set E} {x₀ : E} {V : Submodule ℝ E} (himg : intrinsicInterior ℝ C = x₀ +ᵥ ⇑V.subtype '' interior (chart C x₀ V)) (hC' : C' ⊆ intrinsicInterior ℝ C) (hdense : intrinsicInterior ℝ C ⊆ closure C') :
interior (chart C x₀ V) ⊆ closure (chart C' x₀ V)

Density transports to the chart: if C' is dense in ri C, its chart is dense in the interior of the chart of C.

Pointwise boundedness in the ri form #

theorem Tdaf.ConvexAnalysis.exists_forall_abs_le_of_isCompact_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {ι : Type u_2} {C C' : Set E} {f : ι → E → ℝ} (hC : Convex ℝ C) (hf : ∀ (i : ι), ConvexOn ℝ C (f i)) (hC' : C' ⊆ intrinsicInterior ℝ C) (hdense : intrinsicInterior ℝ C ⊆ closure C') (hab : ∀ x ∈ C', BddAbove (Set.range fun (i : ι) => f i x)) (hbe : ∃ z ∈ intrinsicInterior ℝ C, BddBelow (Set.range fun (i : ι) => f i z)) {S : Set E} (hS : IsCompact S) (hSC : S ⊆ intrinsicInterior ℝ C) :
∃ (M : ℝ), 0 ≤ M ∧ ∀ (i : ι), ∀ x ∈ S, |f i x| ≤ M

The uniform boundedness half: a family of functions convex on a convex set C, pointwise bounded above on a subset C' of ri C whose closure contains ri C and bounded below at one point of ri C, is uniformly bounded on every compact subset of ri C. Taking C' = ri C gives the usual hypothesis, for a relatively open C.

theorem Tdaf.ConvexAnalysis.exists_forall_lipschitzOnWith_of_isCompact_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {ι : Type u_2} {C C' : Set E} {f : ι → E → ℝ} (hC : Convex ℝ C) (hf : ∀ (i : ι), ConvexOn ℝ C (f i)) (hC' : C' ⊆ intrinsicInterior ℝ C) (hdense : intrinsicInterior ℝ C ⊆ closure C') (hab : ∀ x ∈ C', BddAbove (Set.range fun (i : ι) => f i x)) (hbe : ∃ z ∈ intrinsicInterior ℝ C, BddBelow (Set.range fun (i : ι) => f i z)) {S : Set E} (hS : IsCompact S) (hSC : S ⊆ intrinsicInterior ℝ C) :
∃ (K : NNReal), ∀ (i : ι), LipschitzOnWith K (f i) S

The equi-Lipschitz half: under the hypotheses of exists_forall_abs_le_of_isCompact_relint a single Lipschitz constant serves the whole family on any compact subset of ri C.

Convergence, in the ri form #

theorem Tdaf.ConvexAnalysis.exists_tendstoUniformlyOn_of_dense_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {C C' : Set E} {f : ℕ → E → ℝ} (hC : Convex ℝ C) (hf : ∀ (i : ℕ), ConvexOn ℝ C (f i)) (hC' : C' ⊆ intrinsicInterior ℝ C) (hdense : intrinsicInterior ℝ C ⊆ closure C') (hcv : ∀ x ∈ C', ∃ (L : ℝ), Filter.Tendsto (fun (i : ℕ) => f i x) Filter.atTop (nhds L)) :
∃ (g : E → ℝ), ConvexOn ℝ (intrinsicInterior ℝ C) g ∧ (∀ x ∈ intrinsicInterior ℝ C, Filter.Tendsto (fun (i : ℕ) => f i x) Filter.atTop (nhds (g x))) ∧ ∀ ⦃S : Set E⦄, IsCompact S → S ⊆ intrinsicInterior ℝ C → TendstoUniformlyOn f g Filter.atTop S

Convergence from a dense subset: a sequence of functions convex on a convex set C which converges pointwise on a subset C' of ri C whose closure contains ri C converges pointwise on all of ri C, to a finite convex limit, uniformly on every compact subset of ri C.

theorem Tdaf.ConvexAnalysis.tendstoUniformlyOn_of_tendsto_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {C : Set E} {f : ℕ → E → ℝ} (hC : Convex ℝ C) (hf : ∀ (i : ℕ), ConvexOn ℝ C (f i)) {g : E → ℝ} (hg : ∀ x ∈ intrinsicInterior ℝ C, Filter.Tendsto (fun (i : ℕ) => f i x) Filter.atTop (nhds (g x))) {S : Set E} (hS : IsCompact S) (hSC : S ⊆ intrinsicInterior ℝ C) :

The same with the limit supplied: pointwise convergence on ri C upgrades to uniform convergence on its compact subsets.

theorem Tdaf.ConvexAnalysis.eventually_forall_le_add_of_eventually_le_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {C : Set E} {f : ℕ → E → ℝ} (hC : Convex ℝ C) (hf : ∀ (i : ℕ), ConvexOn ℝ C (f i)) {g : E → ℝ} (hg : ConvexOn ℝ C g) (hle : ∀ x ∈ intrinsicInterior ℝ C, ∀ δ > 0, ∀ᶠ (i : ℕ) in Filter.atTop, f i x ≤ g x + δ) {S : Set E} (hS : IsCompact S) (hSC : S ⊆ intrinsicInterior ℝ C) {ε : ℝ} (hε : 0 < ε) :
∀ᶠ (i : ℕ) in Filter.atTop, ∀ x ∈ S, f i x ≤ g x + ε

An eventual upper bound: if limsup_i f i x ≤ g x for every x ∈ ri C, with g convex, then on each compact S ⊆ ri C the bound f i ≤ g + ε holds uniformly for large i.

The limsup hypothesis is spelled as "for every δ > 0, eventually f i x ≤ g x + δ".

theorem Tdaf.ConvexAnalysis.exists_subseq_tendstoUniformlyOn_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {C C' : Set E} {f : ℕ → E → ℝ} (hC : Convex ℝ C) (hf : ∀ (i : ℕ), ConvexOn ℝ C (f i)) (hC' : C' ⊆ intrinsicInterior ℝ C) (hdense : intrinsicInterior ℝ C ⊆ closure C') (hbdd : ∀ x ∈ C', Bornology.IsBounded (Set.range fun (i : ℕ) => f i x)) :
∃ (φ : ℕ → ℕ) (g : E → ℝ), StrictMono φ ∧ ConvexOn ℝ (intrinsicInterior ℝ C) g ∧ (∀ x ∈ intrinsicInterior ℝ C, Filter.Tendsto (fun (i : ℕ) => f (φ i) x) Filter.atTop (nhds (g x))) ∧ ∀ ⦃S : Set E⦄, IsCompact S → S ⊆ intrinsicInterior ℝ C → TendstoUniformlyOn (fun (i : ℕ) => f (φ i)) g Filter.atTop S

Arzelà–Ascoli for convex functions: a sequence of functions convex on C whose values are bounded at each point of a subset C' of ri C with ri C ⊆ cl C' has a subsequence converging uniformly on the compact subsets of ri C to a finite convex function.

Joint continuity, in the ri form #

theorem Tdaf.ConvexAnalysis.continuousOn_prod_of_convexOn_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {C C' : Set E} {T : Type u_2} [TopologicalSpace T] [LocallyCompactSpace T] (hC : Convex ℝ C) {F : E × T → ℝ} (hconv : ∀ (t : T), ConvexOn ℝ C fun (x : E) => F (x, t)) (hC' : C' ⊆ intrinsicInterior ℝ C) (hdense : intrinsicInterior ℝ C ⊆ closure C') (hcont : ∀ x ∈ C', Continuous fun (t : T) => F (x, t)) :

Joint continuity: a real-valued function on ri C × T, with T locally compact, convex in its first argument and continuous in its second, is jointly continuous relative to ri C × T.

The headline form #

theorem Tdaf.ConvexAnalysis.exists_forall_abs_le_and_lipschitzOnWith_of_isCompact_relint {E : Type u_1} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E] {ι : Type u_2} {C : Set E} {f : ι → E → ℝ} (hC : Convex ℝ C) (hf : ∀ (i : ι), ConvexOn ℝ C (f i)) (hbdd : ∀ x ∈ intrinsicInterior ℝ C, Bornology.IsBounded (Set.range fun (i : ι) => f i x)) {S : Set E} (hS : IsCompact S) (hSC : S ⊆ intrinsicInterior ℝ C) :
(∃ (M : ℝ), 0 ≤ M ∧ ∀ (i : ι), ∀ x ∈ S, |f i x| ≤ M) ∧ ∃ (K : NNReal), ∀ (i : ι), LipschitzOnWith K (f i) S

A family of convex functions finite and pointwise bounded on ri C is uniformly bounded and equi-Lipschitzian on every compact subset of ri C.