Uniform integrability #
THIS FILE IS SYNCHRONIZED WITH MATHLIB4. Any changes to this file require a corresponding PR to mathlib4.
This file contains the definitions for uniform integrability (both in the measure theory sense as well as the probability theory sense). This file also contains the Vitali convergence theorem which estabishes a relation between uniform integrability, convergence in measure and Lp convergence.
Uniform integrability plays a vital role in the theory of martingales most notably is used to fomulate the martingale convergence theorem.
Main definitions #
measure_theory.unif_integrable
: uniform integrability in the measure theory sense. In particular, a sequence of functionsf
is uniformly integrable if for allε > 0
, there exists someδ > 0
such that for all setss
of smaller measure thanδ
, the Lp-norm off i
restricteds
is smaller thanε
for alli
.measure_theory.uniform_integrable
: uniform integrability in the probability theory sense. In particular, a sequence of measurable functionsf
is uniformly integrable in the probability theory sense if it is uniformly integrable in the measure theory sense and has uniformly bounded Lp-norm.
Main results #
measure_theory.unif_integrable_fintype
: a finite sequence of Lp functions is uniformly integrable.measure_theory.tendsto_Lp_of_tendsto_ae
: a sequence of Lp functions which is uniformly integrable converges in Lp if they converge almost everywhere.measure_theory.tendsto_in_measure_iff_tendsto_Lp
: Vitali convergence theorem: a sequence of Lp functions converges in Lp if and only if it is uniformly integrable and converges in measure.
Tags #
uniform integrable, uniformly absolutely continuous integral, Vitali convergence theorem
Uniform integrability in the measure theory sense.
A sequence of functions f
is said to be uniformly integrable if for all ε > 0
, there exists
some δ > 0
such that for all sets s
with measure less than δ
, the Lp-norm of f i
restricted on s
is less than ε
.
Uniform integrablility is also known as uniformly absolutely continuous integrals.
Equations
- measure_theory.unif_integrable f p μ = ∀ ⦃ε : ℝ⦄, 0 < ε → (∃ (δ : ℝ) (hδ : 0 < δ), ∀ (i : ι) (s : set α), measurable_set s → ⇑μ s ≤ ennreal.of_real δ → measure_theory.snorm (s.indicator (f i)) p μ ≤ ennreal.of_real ε)
In probability theory, a family of measurable functions is uniformly integrable if it is uniformly integrable in the measure theory sense and is uniformly bounded.
Equations
- measure_theory.uniform_integrable f p μ = ((∀ (i : ι), measure_theory.ae_strongly_measurable (f i) μ) ∧ measure_theory.unif_integrable f p μ ∧ ∃ (C : nnreal), ∀ (i : ι), measure_theory.snorm (f i) p μ ≤ ↑C)
unif_integrable
#
This section deals with uniform integrability in the measure theory sense.
This lemma is weaker than measure_theory.mem_ℒp.integral_indicator_norm_ge_nonneg_le
as the latter provides 0 ≤ M
and does not require the measurability of f
.
This lemma is superceded by measure_theory.mem_ℒp.integral_indicator_norm_ge_nonneg_le
which does not require measurability.
This lemma implies that a single function is uniformly integrable (in the probability sense).
Auxiliary lemma for measure_theory.mem_ℒp.snorm_indicator_le
.
This lemma is superceded by measure_theory.mem_ℒp.snorm_indicator_le
which does not require
measurability on f
.
A constant function is uniformly integrable.
A single function is uniformly integrable.
This lemma is less general than measure_theory.unif_integrable_fintype
which applies to
all sequences indexed by a finite type.
A finite sequence of Lp functions is uniformly integrable.
A sequence of uniformly integrable functions which converges μ-a.e. converges in Lp.
A sequence of uniformly integrable functions which converges μ-a.e. converges in Lp.
Convergence in Lp implies uniform integrability.
Forward direction of Vitali's convergence theorem: if f
is a sequence of uniformly integrable
functions that converge in measure to some function g
in a finite measure space, then f
converge in Lp to g
.
Vitali's convergence theorem: A sequence of functions f
converges to g
in Lp if and
only if it is uniformly integrable and converges to g
in measure.
This lemma is superceded by unif_integrable_of
which do not require C
to be positive.
uniform_integrable
In probability theory, uniform integrability normally refers to the condition that a sequence
of function (fₙ)
satisfies for all ε > 0
, there exists some C ≥ 0
such that
∫ x in {|fₙ| ≥ C}, fₙ x ∂μ ≤ ε
for all n
.
In this section, we will develope some API for uniform_integrable
and prove that
uniform_integrable
is equivalent to this definition of uniform integrability.
A finite sequence of Lp functions is uniformly integrable in the probability sense.
A single function is uniformly integrable in the probability sense.
A constant sequence of functions is uniformly integrable in the probability sense.
This lemma is superceded by uniform_integrable_of
which only requires
ae_strongly_measurable
.
A sequene of functions (fₙ)
is uniformly integrable in the probability sense if for all
ε > 0
, there exists some C
such that ∫ x in {|fₙ| ≥ C}, fₙ x ∂μ ≤ ε
for all n
.
This lemma is superceded by uniform_integrable.spec
which does not require measurability.
The definition of uniform integrable in mathlib is equivalent to the definition commonly found in literature.
The averaging of a uniformly integrable sequence is also uniformly integrable.