Datasets:
Tasks:
Text Generation
Modalities:
Text
Sub-tasks:
language-modeling
Languages:
English
Size:
100K - 1M
License:
File size: 9,336 Bytes
4365a98 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 |
/-
Copyright (c) 2021 Yaël Dillies, Bhavik Mehta. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Yaël Dillies, Bhavik Mehta
-/
import analysis.convex.extreme
import analysis.convex.function
import analysis.normed_space.ordered
/-!
# Exposed sets
This file defines exposed sets and exposed points for sets in a real vector space.
An exposed subset of `A` is a subset of `A` that is the set of all maximal points of a functional
(a continuous linear map `E → 𝕜`) over `A`. By convention, `∅` is an exposed subset of all sets.
This allows for better functoriality of the definition (the intersection of two exposed subsets is
exposed, faces of a polytope form a bounded lattice).
This is an analytic notion of "being on the side of". It is stronger than being extreme (see
`is_exposed.is_extreme`), but weaker (for exposed points) than being a vertex.
An exposed set of `A` is sometimes called a "face of `A`", but we decided to reserve this
terminology to the more specific notion of a face of a polytope (sometimes hopefully soon out
on mathlib!).
## Main declarations
* `is_exposed 𝕜 A B`: States that `B` is an exposed set of `A` (in the literature, `A` is often
implicit).
* `is_exposed.is_extreme`: An exposed set is also extreme.
## References
See chapter 8 of [Barry Simon, *Convexity*][simon2011]
## TODO
Define intrinsic frontier/interior and prove the lemmas related to exposed sets and points.
Generalise to Locally Convex Topological Vector Spaces™
More not-yet-PRed stuff is available on the branch `sperner_again`.
-/
open_locale classical affine big_operators
open set
variables (𝕜 : Type*) {E : Type*} [normed_linear_ordered_field 𝕜] [normed_add_comm_group E]
[normed_space 𝕜 E] {l : E →L[𝕜] 𝕜} {A B C : set E} {X : finset E} {x : E}
/-- A set `B` is exposed with respect to `A` iff it maximizes some functional over `A` (and contains
all points maximizing it). Written `is_exposed 𝕜 A B`. -/
def is_exposed (A B : set E) : Prop :=
B.nonempty → ∃ l : E →L[𝕜] 𝕜, B = {x ∈ A | ∀ y ∈ A, l y ≤ l x}
variables {𝕜}
/-- A useful way to build exposed sets from intersecting `A` with halfspaces (modelled by an
inequality with a functional). -/
def continuous_linear_map.to_exposed (l : E →L[𝕜] 𝕜) (A : set E) : set E :=
{x ∈ A | ∀ y ∈ A, l y ≤ l x}
lemma continuous_linear_map.to_exposed.is_exposed : is_exposed 𝕜 A (l.to_exposed A) := λ h, ⟨l, rfl⟩
lemma is_exposed_empty : is_exposed 𝕜 A ∅ :=
λ ⟨x, hx⟩, by { exfalso, exact hx }
namespace is_exposed
protected lemma subset (hAB : is_exposed 𝕜 A B) : B ⊆ A :=
begin
rintro x hx,
obtain ⟨_, rfl⟩ := hAB ⟨x, hx⟩,
exact hx.1,
end
@[refl] protected lemma refl (A : set E) : is_exposed 𝕜 A A :=
λ ⟨w, hw⟩, ⟨0, subset.antisymm (λ x hx, ⟨hx, λ y hy, by exact le_refl 0⟩) (λ x hx, hx.1)⟩
protected lemma antisymm (hB : is_exposed 𝕜 A B) (hA : is_exposed 𝕜 B A) :
A = B :=
hA.subset.antisymm hB.subset
/- `is_exposed` is *not* transitive: Consider a (topologically) open cube with vertices
`A₀₀₀, ..., A₁₁₁` and add to it the triangle `A₀₀₀A₀₀₁A₀₁₀`. Then `A₀₀₁A₀₁₀` is an exposed subset
of `A₀₀₀A₀₀₁A₀₁₀` which is an exposed subset of the cube, but `A₀₀₁A₀₁₀` is not itself an exposed
subset of the cube. -/
protected lemma mono (hC : is_exposed 𝕜 A C) (hBA : B ⊆ A) (hCB : C ⊆ B) :
is_exposed 𝕜 B C :=
begin
rintro ⟨w, hw⟩,
obtain ⟨l, rfl⟩ := hC ⟨w, hw⟩,
exact ⟨l, subset.antisymm (λ x hx, ⟨hCB hx, λ y hy, hx.2 y (hBA hy)⟩)
(λ x hx, ⟨hBA hx.1, λ y hy, (hw.2 y hy).trans (hx.2 w (hCB hw))⟩)⟩,
end
/-- If `B` is an exposed subset of `A`, then `B` is the intersection of `A` with some closed
halfspace. The converse is *not* true. It would require that the corresponding open halfspace
doesn't intersect `A`. -/
lemma eq_inter_halfspace (hAB : is_exposed 𝕜 A B) :
∃ l : E →L[𝕜] 𝕜, ∃ a, B = {x ∈ A | a ≤ l x} :=
begin
obtain hB | hB := B.eq_empty_or_nonempty,
{ refine ⟨0, 1, _⟩,
rw [hB, eq_comm, eq_empty_iff_forall_not_mem],
rintro x ⟨-, h⟩,
rw continuous_linear_map.zero_apply at h,
linarith },
obtain ⟨l, rfl⟩ := hAB hB,
obtain ⟨w, hw⟩ := hB,
exact ⟨l, l w, subset.antisymm (λ x hx, ⟨hx.1, hx.2 w hw.1⟩)
(λ x hx, ⟨hx.1, λ y hy, (hw.2 y hy).trans hx.2⟩)⟩,
end
protected lemma inter (hB : is_exposed 𝕜 A B) (hC : is_exposed 𝕜 A C) :
is_exposed 𝕜 A (B ∩ C) :=
begin
rintro ⟨w, hwB, hwC⟩,
obtain ⟨l₁, rfl⟩ := hB ⟨w, hwB⟩,
obtain ⟨l₂, rfl⟩ := hC ⟨w, hwC⟩,
refine ⟨l₁ + l₂, subset.antisymm _ _⟩,
{ rintro x ⟨⟨hxA, hxB⟩, ⟨-, hxC⟩⟩,
exact ⟨hxA, λ z hz, add_le_add (hxB z hz) (hxC z hz)⟩ },
rintro x ⟨hxA, hx⟩,
refine ⟨⟨hxA, λ y hy, _⟩, hxA, λ y hy, _⟩,
{ exact (add_le_add_iff_right (l₂ x)).1 ((add_le_add (hwB.2 y hy) (hwC.2 x hxA)).trans
(hx w hwB.1)) },
{ exact (add_le_add_iff_left (l₁ x)).1 (le_trans (add_le_add (hwB.2 x hxA) (hwC.2 y hy))
(hx w hwB.1)) }
end
lemma sInter {F : finset (set E)} (hF : F.nonempty)
(hAF : ∀ B ∈ F, is_exposed 𝕜 A B) :
is_exposed 𝕜 A (⋂₀ F) :=
begin
revert hF F,
refine finset.induction _ _,
{ rintro h,
exfalso,
exact empty_not_nonempty h },
rintro C F _ hF _ hCF,
rw [finset.coe_insert, sInter_insert],
obtain rfl | hFnemp := F.eq_empty_or_nonempty,
{ rw [finset.coe_empty, sInter_empty, inter_univ],
exact hCF C (finset.mem_singleton_self C) },
exact (hCF C (finset.mem_insert_self C F)).inter (hF hFnemp (λ B hB,
hCF B(finset.mem_insert_of_mem hB))),
end
lemma inter_left (hC : is_exposed 𝕜 A C) (hCB : C ⊆ B) :
is_exposed 𝕜 (A ∩ B) C :=
begin
rintro ⟨w, hw⟩,
obtain ⟨l, rfl⟩ := hC ⟨w, hw⟩,
exact ⟨l, subset.antisymm (λ x hx, ⟨⟨hx.1, hCB hx⟩, λ y hy, hx.2 y hy.1⟩)
(λ x ⟨⟨hxC, _⟩, hx⟩, ⟨hxC, λ y hy, (hw.2 y hy).trans (hx w ⟨hC.subset hw, hCB hw⟩)⟩)⟩,
end
lemma inter_right (hC : is_exposed 𝕜 B C) (hCA : C ⊆ A) :
is_exposed 𝕜 (A ∩ B) C :=
begin
rw inter_comm,
exact hC.inter_left hCA,
end
protected lemma is_extreme (hAB : is_exposed 𝕜 A B) :
is_extreme 𝕜 A B :=
begin
refine ⟨hAB.subset, λ x₁ hx₁A x₂ hx₂A x hxB hx, _⟩,
obtain ⟨l, rfl⟩ := hAB ⟨x, hxB⟩,
have hl : convex_on 𝕜 univ l := l.to_linear_map.convex_on convex_univ,
have hlx₁ := hxB.2 x₁ hx₁A,
have hlx₂ := hxB.2 x₂ hx₂A,
refine ⟨⟨hx₁A, λ y hy, _⟩, ⟨hx₂A, λ y hy, _⟩⟩,
{ rw hlx₁.antisymm (hl.le_left_of_right_le (mem_univ _) (mem_univ _) hx hlx₂),
exact hxB.2 y hy },
{ rw hlx₂.antisymm (hl.le_right_of_left_le (mem_univ _) (mem_univ _) hx hlx₁),
exact hxB.2 y hy }
end
protected lemma convex (hAB : is_exposed 𝕜 A B) (hA : convex 𝕜 A) :
convex 𝕜 B :=
begin
obtain rfl | hB := B.eq_empty_or_nonempty,
{ exact convex_empty },
obtain ⟨l, rfl⟩ := hAB hB,
exact λ x₁ x₂ hx₁ hx₂ a b ha hb hab, ⟨hA hx₁.1 hx₂.1 ha hb hab, λ y hy,
((l.to_linear_map.concave_on convex_univ).convex_ge _
⟨mem_univ _, hx₁.2 y hy⟩ ⟨mem_univ _, hx₂.2 y hy⟩ ha hb hab).2⟩,
end
protected lemma is_closed [order_closed_topology 𝕜] (hAB : is_exposed 𝕜 A B) (hA : is_closed A) :
is_closed B :=
begin
obtain ⟨l, a, rfl⟩ := hAB.eq_inter_halfspace,
exact hA.is_closed_le continuous_on_const l.continuous.continuous_on,
end
protected lemma is_compact [order_closed_topology 𝕜] (hAB : is_exposed 𝕜 A B) (hA : is_compact A) :
is_compact B :=
compact_of_is_closed_subset hA (hAB.is_closed hA.is_closed) hAB.subset
end is_exposed
variables (𝕜)
/-- A point is exposed with respect to `A` iff there exists an hyperplane whose intersection with
`A` is exactly that point. -/
def set.exposed_points (A : set E) :
set E :=
{x ∈ A | ∃ l : E →L[𝕜] 𝕜, ∀ y ∈ A, l y ≤ l x ∧ (l x ≤ l y → y = x)}
variables {𝕜}
lemma exposed_point_def :
x ∈ A.exposed_points 𝕜 ↔ x ∈ A ∧ ∃ l : E →L[𝕜] 𝕜, ∀ y ∈ A, l y ≤ l x ∧ (l x ≤ l y → y = x) :=
iff.rfl
lemma exposed_points_subset :
A.exposed_points 𝕜 ⊆ A :=
λ x hx, hx.1
@[simp] lemma exposed_points_empty :
(∅ : set E).exposed_points 𝕜 = ∅ :=
subset_empty_iff.1 exposed_points_subset
/-- Exposed points exactly correspond to exposed singletons. -/
lemma mem_exposed_points_iff_exposed_singleton :
x ∈ A.exposed_points 𝕜 ↔ is_exposed 𝕜 A {x} :=
begin
use λ ⟨hxA, l, hl⟩ h, ⟨l, eq.symm $ eq_singleton_iff_unique_mem.2 ⟨⟨hxA, λ y hy, (hl y hy).1⟩,
λ z hz, (hl z hz.1).2 (hz.2 x hxA)⟩⟩,
rintro h,
obtain ⟨l, hl⟩ := h ⟨x, mem_singleton _⟩,
rw [eq_comm, eq_singleton_iff_unique_mem] at hl,
exact ⟨hl.1.1, l, λ y hy, ⟨hl.1.2 y hy, λ hxy, hl.2 y ⟨hy, λ z hz, (hl.1.2 z hz).trans hxy⟩⟩⟩,
end
lemma exposed_points_subset_extreme_points :
A.exposed_points 𝕜 ⊆ A.extreme_points 𝕜 :=
λ x hx, mem_extreme_points_iff_extreme_singleton.2
(mem_exposed_points_iff_exposed_singleton.1 hx).is_extreme
|