Outer Measures #
An outer measure is a function μ : Set α → ℝ≥0∞
, from the powerset of a type to the extended
nonnegative real numbers that satisfies the following conditions:
μ ∅ = 0
;μ
is monotone;μ
is countably subadditive. This means that the outer measure of a countable union is at most the sum of the outer measure on the individual sets.
Note that we do not need α
to be measurable to define an outer measure.
The outer measures on a type α
form a complete lattice.
Given an arbitrary function m : Set α → ℝ≥0∞
that sends ∅
to 0
we can define an outer
measure on α
that on s
is defined to be the infimum of ∑ᵢ, m (sᵢ)
for all collections of sets
sᵢ
that cover s
. This is the unique maximal outer measure that is at most the given function.
We also define this for functions m
defined on a subset of Set α
, by treating the function as
having value ∞
outside its domain.
Given an outer measure m
, the Carathéodory-measurable sets are the sets s
such that
for all sets t
we have m t = m (t ∩ s) + m (t \ s)
. This forms a measurable space.
Main definitions and statements #
OuterMeasure.boundedBy
is the greatest outer measure that is at most the given function. If you know that the given function sends∅
to0
, thenOuterMeasure.ofFunction
is a special case.caratheodory
is the Carathéodory-measurable space of an outer measure.sInf_eq_boundedBy_sInfGen
is a characterization of the infimum of outer measures.inducedOuterMeasure
is the measure induced by a function on a subset ofSet α
References #
Tags #
outer measure, Carathéodory-measurable, Carathéodory's criterion
An outer measure is a countably subadditive monotone function that sends ∅
to 0
.
Instances For
Equations
- MeasureTheory.OuterMeasure.instCoeFun = { coe := fun (m : MeasureTheory.OuterMeasure α) => ↑m }
Alias of the reverse direction of MeasureTheory.OuterMeasure.iUnion_null_iff
.
If a set has zero measure in a neighborhood of each of its points, then it has zero measure in a second-countable space.
If m s ≠ 0
, then for some point x ∈ s
and any t ∈ 𝓝[s] x
we have 0 < m t
.
If s : ι → Set α
is a sequence of sets, S = ⋃ n, s n
, and m (S \ s n)
tends to zero along
some nontrivial filter (usually atTop
on ι = ℕ
), then m S = ⨆ n, m (s n)
.
If s : ℕ → Set α
is a monotone sequence of sets such that ∑' k, m (s (k + 1) \ s k) ≠ ∞
,
then m (⋃ n, s n) = ⨆ n, m (s n)
.
A version of MeasureTheory.OuterMeasure.ext
that assumes μ₁ s = μ₂ s
on all nonempty
sets s
, and gets μ₁ ∅ = μ₂ ∅
from MeasureTheory.OuterMeasure.empty'
.
Equations
- MeasureTheory.OuterMeasure.instInhabited = { default := 0 }
Equations
- One or more equations did not get rendered due to their size.
Equations
- One or more equations did not get rendered due to their size.
Equations
- ⋯ = ⋯
Equations
- ⋯ = ⋯
Equations
- ⋯ = ⋯
Equations
- MeasureTheory.OuterMeasure.instMulAction = Function.Injective.mulAction (fun (μ : MeasureTheory.OuterMeasure α) (s : Set α) => ↑μ s) ⋯ ⋯
Equations
- MeasureTheory.OuterMeasure.addCommMonoid = Function.Injective.addCommMonoid (let_fun this := fun (μ : MeasureTheory.OuterMeasure α) (s : Set α) => ↑μ s; this) ⋯ ⋯ ⋯ ⋯
(⇑)
as an AddMonoidHom
.
Equations
- MeasureTheory.OuterMeasure.coeFnAddMonoidHom = { toZeroHom := { toFun := MeasureTheory.OuterMeasure.measureOf, map_zero' := ⋯ }, map_add' := ⋯ }
Instances For
Equations
- MeasureTheory.OuterMeasure.instDistribMulAction = Function.Injective.distribMulAction MeasureTheory.OuterMeasure.coeFnAddMonoidHom ⋯ ⋯
Equations
- MeasureTheory.OuterMeasure.instModule = Function.Injective.module R MeasureTheory.OuterMeasure.coeFnAddMonoidHom ⋯ ⋯
Equations
- MeasureTheory.OuterMeasure.instBot = { bot := 0 }
Equations
- MeasureTheory.OuterMeasure.instPartialOrder = PartialOrder.mk ⋯
Equations
- MeasureTheory.OuterMeasure.orderBot = OrderBot.mk ⋯
Equations
- One or more equations did not get rendered due to their size.
Equations
- One or more equations did not get rendered due to their size.
The pushforward of m
along f
. The outer measure on s
is defined to be m (f ⁻¹' s)
.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
The dirac outer measure.
Equations
- MeasureTheory.OuterMeasure.dirac a = { measureOf := fun (s : Set α) => Set.indicator s (fun (x : α) => 1) a, empty := ⋯, mono := ⋯, iUnion_nat := ⋯ }
Instances For
The sum of an (arbitrary) collection of outer measures.
Equations
- MeasureTheory.OuterMeasure.sum f = { measureOf := fun (s : Set α) => ∑' (i : ι), ↑(f i) s, empty := ⋯, mono := ⋯, iUnion_nat := ⋯ }
Instances For
Pullback of an OuterMeasure
: comap f μ s = μ (f '' s)
.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Restrict an OuterMeasure
to a set.
Equations
- MeasureTheory.OuterMeasure.restrict s = LinearMap.comp (MeasureTheory.OuterMeasure.map Subtype.val) (MeasureTheory.OuterMeasure.comap Subtype.val)
Instances For
Given any function m
assigning measures to sets satisying m ∅ = 0
, there is
a unique maximal outer measure μ
satisfying μ s ≤ m s
for all s : Set α
.
Equations
- One or more equations did not get rendered due to their size.
Instances For
If m u = ∞
for any set u
that has nonempty intersection both with s
and t
, then
μ (s ∪ t) = μ s + μ t
, where μ = MeasureTheory.OuterMeasure.ofFunction m m_empty
.
E.g., if α
is an (e)metric space and m u = ∞
on any set of diameter ≥ r
, then this lemma
implies that μ (s ∪ t) = μ s + μ t
on any two sets such that r ≤ edist x y
for all x ∈ s
and y ∈ t
.
Given any function m
assigning measures to sets, there is a unique maximal outer measure μ
satisfying μ s ≤ m s
for all s : Set α
. This is the same as OuterMeasure.ofFunction
,
except that it doesn't require m ∅ = 0
.
Equations
- MeasureTheory.OuterMeasure.boundedBy m = MeasureTheory.OuterMeasure.ofFunction (fun (s : Set α) => ⨆ (_ : Set.Nonempty s), m s) ⋯
Instances For
If m u = ∞
for any set u
that has nonempty intersection both with s
and t
, then
μ (s ∪ t) = μ s + μ t
, where μ = MeasureTheory.OuterMeasure.boundedBy m
.
E.g., if α
is an (e)metric space and m u = ∞
on any set of diameter ≥ r
, then this lemma
implies that μ (s ∪ t) = μ s + μ t
on any two sets such that r ≤ edist x y
for all x ∈ s
and y ∈ t
.
A set s
is Carathéodory-measurable for an outer measure m
if for all sets t
we have
m t = m (t ∩ s) + m (t \ s)
.
Equations
Instances For
The Carathéodory-measurable sets for an outer measure m
form a Dynkin system.
Equations
- MeasureTheory.OuterMeasure.caratheodoryDynkin m = { Has := MeasureTheory.OuterMeasure.IsCaratheodory m, has_empty := ⋯, has_compl := ⋯, has_iUnion_nat := ⋯ }
Instances For
Given an outer measure μ
, the Carathéodory-measurable space is
defined such that s
is measurable if ∀t, μ t = μ (t ∩ s) + μ (t \ s)
.
Equations
Instances For
Given a set of outer measures, we define a new function that on a set s
is defined to be the
infimum of μ(s)
for the outer measures μ
in the collection. We ensure that this
function is defined to be 0
on ∅
, even if the collection of outer measures is empty.
The outer measure generated by this function is the infimum of the given outer measures.
Equations
- MeasureTheory.OuterMeasure.sInfGen m s = ⨅ μ ∈ m, ↑μ s
Instances For
The value of the Infimum of a nonempty set of outer measures on a set is not simply the minimum value of a measure on that set: it is the infimum sum of measures of countable set of sets that covers that set, where a different measure can be used for each set in the cover.
The value of the Infimum of a set of outer measures on a nonempty set is not simply the minimum value of a measure on that set: it is the infimum sum of measures of countable set of sets that covers that set, where a different measure can be used for each set in the cover.
The value of the Infimum of a nonempty family of outer measures on a set is not simply the minimum value of a measure on that set: it is the infimum sum of measures of countable set of sets that covers that set, where a different measure can be used for each set in the cover.
The value of the Infimum of a family of outer measures on a nonempty set is not simply the minimum value of a measure on that set: it is the infimum sum of measures of countable set of sets that covers that set, where a different measure can be used for each set in the cover.
The value of the Infimum of a nonempty family of outer measures on a set is not simply the minimum value of a measure on that set: it is the infimum sum of measures of countable set of sets that covers that set, where a different measure can be used for each set in the cover.
The value of the Infimum of a nonempty family of outer measures on a set is not simply the minimum value of a measure on that set: it is the infimum sum of measures of countable set of sets that covers that set, where a different measure can be used for each set in the cover.
This proves that Inf and restrict commute for outer measures, so long as the set of outer measures is nonempty.
Induced Outer Measure #
We can extend a function defined on a subset of Set α
to an outer measure.
The underlying function is called extend
, and the measure it induces is called
inducedOuterMeasure
.
Some lemmas below are proven twice, once in the general case, and one where the function m
is only defined on measurable sets (i.e. when P = MeasurableSet
). In the latter cases, we can
remove some hypotheses in the statement. The general version has the same name, but with a prime
at the end.
We can trivially extend a function defined on a subclass of objects (with codomain ℝ≥0∞
)
to all objects by defining it to be ∞
on the objects not in the class.
Equations
- MeasureTheory.extend m s = ⨅ (h : P s), m s h
Instances For
Given an arbitrary function on a subset of sets, we can define the outer measure corresponding
to it (this is the unique maximal outer measure that is at most m
on the domain of m
).
Equations
Instances For
If P u
is False
for any set u
that has nonempty intersection both with s
and t
, then
μ (s ∪ t) = μ s + μ t
, where μ = inducedOuterMeasure m P0 m0
.
E.g., if α
is an (e)metric space and P u = diam u < r
, then this lemma implies that
μ (s ∪ t) = μ s + μ t
on any two sets such that r ≤ edist x y
for all x ∈ s
and y ∈ t
.
To test whether s
is Carathéodory-measurable we only need to check the sets t
for which
P t
holds. See ofFunction_caratheodory
for another way to show the Carathéodory-measurability
of s
.
If P
is MeasurableSet
for some measurable space, then we can remove some hypotheses of the
above lemmas.
Given an outer measure m
we can forget its value on non-measurable sets, and then consider
m.trim
, the unique maximal outer measure less than that function.
Equations
- MeasureTheory.OuterMeasure.trim m = MeasureTheory.inducedOuterMeasure (fun (s : Set α) (x : MeasurableSet s) => ↑m s) ⋯ ⋯
Instances For
If μ i
is a countable family of outer measures, then for every set s
there exists
a measurable set t ⊇ s
such that μ i t = (μ i).trim s
for all i
.
If m₁ s = op (m₂ s) (m₃ s)
for all s
, then the same is true for m₁.trim
, m₂.trim
,
and m₃ s
.
If m₁ s = op (m₂ s)
for all s
, then the same is true for m₁.trim
and m₂.trim
.
trim
is additive.
trim
respects scalar multiplication.
trim
sends the supremum of two outer measures to the supremum of the trimmed measures.
trim
sends the supremum of a countable family of outer measures to the supremum
of the trimmed measures.
The trimmed property of a measure μ states that μ.toOuterMeasure.trim = μ.toOuterMeasure
.
This theorem shows that a restricted trimmed outer measure is a trimmed outer measure.