PhyslibAlpha.ProbabilisticTheory.Measurement.BoundedIntegral
Integrating bounded functions against an effect-valued measure
i. Overview
If the observables are complete for the order-unit norm, the integral of simple functions extends to bounded measurable functions. A bounded function is uniformly approximated by simple functions on a mesh of width `1 / (n + 1)`, and simple functions within `ε` and `ε'` of `f` have integrals within `ε + ε'`. So the integrals of the approximations converge, to a limit that does not depend on the approximation or the bound. The integral is linear and positive, and agrees with the simple integral on simple functions.
ii. Key results
- `EffectValuedMeasure.meshPiece`, `EffectValuedMeasure.meshWeight` : the mesh approximation. - `EffectValuedMeasure.orderUnitNorm_simpleIntegral_sub_le` : nearby simple functions have nearby integrals. - `EffectValuedMeasure.integral` : the integral of a bounded measurable function. - `EffectValuedMeasure.tendsto_of_uniformly_approximating` : every uniform approximation gives the integral. - `EffectValuedMeasure.integral_add`, `EffectValuedMeasure.nonneg_integral` : linearity and positivity.
iii. Table of contents
- A. The mesh approximation of a bounded measurable function
- B. Linear operations on simple values
- C. Comparing the simple integrals of two approximations
- D. The integral, via completeness
A. The mesh approximation of a bounded measurable function
B. Linear operations on simple values
C. Comparing the simple integrals of two approximations
The key ingredient for both the Cauchy property and the independence of the eventual integral from the choice of approximating sequence: if two simple functions (over possibly different partitions) both stay within `ε`, resp. `ε'`, of the same `f` everywhere, their integrals against `μ` stay within `ε + ε'` of each other, in the order-unit norm.
D. The integral, via completeness
The order-unit norm makes `E` a normed group, assumed complete. Most order-unit spaces are not complete.
40 declarations
Number of mesh cells for bound and mesh parameter
Given a bound and a natural number (where the mesh width is defined as ), this function returns the number of mesh cells required to cover the interval from to . It is defined as the ceiling of the product of and , expressed as .
Index set for the mesh on at scale
Given a bound and a discretization parameter , the type `MeshIndex M n` is the set of integer labels for the mesh cells used to partition the interval . It is defined as the set of integers such that , where is the value determined by `meshBound M n`.
Mesh cell weight at scale
Given a bound , a discretization parameter , and an index from the set of mesh indices , the function returns the real value . This value represents the "weight" of the -th mesh cell, corresponding to the left endpoint of that cell.
The -th mesh piece of at scale :
Given a function , a real bound , a discretization parameter (defining a mesh width of ), and a mesh index , the set `meshPiece f M n k` is the subset of consisting of all points such that . This set is the pullback along of the -th mesh interval, representing the points where lies in the half-open interval .
For a function , a real bound , a discretization parameter , and a mesh index , a point belongs to the -th mesh piece if and only if the floor of is equal to :
Measurability of the Mesh Pieces of
Let be a measurable function. For any discretization parameter and any mesh index , the -th mesh piece defined by the set is a measurable set.
Mesh index for point
Given a function bounded by (such that for all ) and a discretization parameter (where the mesh width is ), this function maps a point to its corresponding mesh cell index . This index is guaranteed to lie within the range of the mesh index set .
The mesh index of is
For a function bounded by , a natural number , and a point , the mesh index of (viewed as an integer) is equal to the floor of the value , expressed as:
Let be a function and be a real number such that for all . For any and any , let be the mesh index of (denoted as `meshIndexOf hM n x`). Then is an element of the -th mesh piece of at scale , which is defined as the set .
The mesh pieces of form a partition of at scale
Let be a function bounded by (i.e., for all ). For any natural number , the collection of mesh pieces , indexed by such that , forms a partition of the space .
The value of the mesh simple function at is
Given a function , a bound , and a discretization parameter , let the -th mesh piece be and the corresponding mesh weight be . For any , the value of the simple function at is given by where denotes the floor function.
Uniform error bound for the mesh approximation
For any function , a real bound , and a discretization parameter , the mesh-based simple function approximation (with mesh width ) evaluated at any point satisfies the inequality: Here, `meshPiece f M n` partitions into sets where the value of falls into the -th mesh interval, and `meshWeight M n` assigns the value to the -th piece. In essence, the mesh approximation stays within of for all .
Additivity of `simpleValue` in the weights
For a finite index set , real coefficients , a family of sets , and a point , the value of the simple function is additive with respect to its coefficients: where is the indicator function of the set .
Scalar Multiplication Homogeneity of the Value of a Simple Function
Let be a finite index set, be a scalar, and be a collection of coefficients. For any family of sets and any point , the value of the simple function with scaled coefficients at is equal to times the value of the simple function with coefficients at . In terms of the indicator function , this states:
The Intersection of Two Partitions is a Partition
Let be a set, and let and be finite index sets. Suppose and are partitions of . Then the collection of pairwise intersections indexed by the product set is also a partition of .
for partitions and
Let and be finite index sets, and let and be partitions of a space . For any coefficients and any point , the value of the simple function defined by coefficients and the partition at is equal to the value of the simple function defined by the same coefficients and the refined partition at . That is: where denotes the indicator function of the set .
Equality of a simple function's value on a partition and its product partition refinement
Let be a space, and let and be partitions of indexed by finite sets and . For any point and any collection of real coefficients , the value of the simple function defined by the coefficients and the partition at is equal to the value of the simple function defined over the product partition using the coefficients from the second index (i.e., ). That is, where denotes the indicator function of the set .
The sum of two simple values is the simple value on their common refinement partition
Let be a set and let and be two partitions of indexed by finite sets and . Let and be coefficients. For any , the sum of the values of the two simple functions at is equal to the value of the simple function defined by the common refinement partition with summed weights: where denotes the indicator function of a set .
Equality of a simple integral and its product partition refinement
Let be a space and be an effect-valued measure on taking values in . Let and be partitions of indexed by finite sets and . For any collection of real coefficients , the simple integral of the function defined by and is equal to the simple integral defined over the product partition with coefficients determined by the first index of the product. Specifically,
The simple integral on a partition is equal to its integral on a refined partition (second component)
Let be a measurable space and be a space of effects. Let be an effect-valued measure on taking values in . Let and be finite index sets. Suppose and are two partitions of . For any collection of real coefficients , the simple integral of the function defined by on the partition is equal to the simple integral defined on the product refinement partition using the coefficients from the second index ():
Difference of simple integrals over a common refinement
Let be a measurable space and be an order-unit space. Let be an effect-valued measure on taking values in . Suppose and are two finite partitions of indexed by finite types and . For any collections of real coefficients and , the difference between their respective simple integrals is equal to the simple integral of the difference of the coefficients over the common refinement partition :
The sum of two simple integrals is a simple integral over their common refinement
Let be a measurable space and be an order-unit space. Let be an effect-valued measure on taking values in . Suppose and are two finite partitions of with corresponding real coefficients and . The sum of their simple integrals is equal to the simple integral over the common refinement of the partitions with weights :
The sum of an effect-valued measure over a partition is
Let be an effect-valued measure on a space with values in an order-unit space . For any finite partition of into measurable sets, the sum of the measures of the sets in the partition is equal to the unit element (the certain outcome) :
The order-unit norm of a simple integral is bounded by the maximum absolute value of its coefficients.
Let be an effect-valued measure on a measurable space with values in an order-unit space . Let be a finite partition of into measurable sets and be coefficients for each . If there exists a constant such that for every where the set is non-empty, then the order-unit norm of the simple integral of the function against is also bounded by : where denotes the order-unit norm in .
Non-negativity of simple integral for locally non-negative coefficients
Let be an effect-valued measure on a space taking values in an order-unit space . For a finite partition of and real coefficients , if for all indices such that is non-empty, then the simple integral of the function against is non-negative:
for simple functions uniformly approximating within
Let be an effect-valued measure on a space with values in an order-unit space equipped with the order-unit norm . Let be a function. Suppose and are two simple functions defined by finite partitions and of with real coefficients and respectively, such that and . If there exist non-negative constants such that and for all , then the difference between their simple integrals against satisfies:
Let be an order-unit space with order unit and order-unit norm . For any element and any real number , if the order-unit norm of is strictly less than (i.e., ), then is bounded above by (i.e., ).
is a Normed Additive Commutative Group under the order-unit norm
Let be an ordered additive commutative group equipped with an Archimedean order unit . The space is a normed additive commutative group under the order-unit norm , where for any , the norm is defined as .
is complete under the order-unit norm
Let be an ordered additive commutative group with an Archimedean order unit . The space is a complete metric space with respect to the order-unit norm , which is defined for any as .
for the order-unit norm
Let be an ordered additive commutative group equipped with an Archimedean order unit. For any two elements , the metric distance between them is equal to the order-unit norm of their difference, that is, .
If , then in the order-unit norm space
Let be an ordered additive commutative group with an Archimedean order unit , equipped with the order-unit norm . If a sequence in converges to a limit (with respect to the topology induced by the norm), then for any scalar , the sequence of scalar multiples converges to .
The Mesh Approximation Sequence of Simple Integrals is Cauchy
Let be an effect-valued measure on a measurable space taking values in a space equipped with an order-unit norm . Let be a measurable function bounded by (i.e., for all ). For each , define the mesh pieces and the weights . The sequence of simple integrals defined by is a Cauchy sequence in with respect to the order-unit norm.
Integral of a bounded measurable function against an effect-valued measure
Let be a complete Archimedean order-unit space equipped with the order-unit norm , and let be an effect-valued measure on a measurable space with values in . Given a measurable function and a real bound such that for all , the integral of against is defined as the limit in of the sequence of simple integrals as . This sequence is defined by a uniform mesh approximation: where the sum is taken over the finite set of indices such that the resulting sets partition . The limit exists because the sequence is Cauchy and is assumed to be complete.
The Mesh Approximation Sequence Converges to the Integral
Let be a complete Archimedean order-unit space and be an effect-valued measure on a measurable space taking values in . Let be a measurable function bounded by (such that for all ). For each , define the mesh pieces and the mesh weights . Then the sequence of simple integrals converges to the integral in the order-unit norm of as .
Uniform Convergence of Simple Functions Implies Convergence of Their Integrality against an Effect-Valued Measure
Let be a measurable space and be a complete Archimedean order-unit space equipped with the order-unit norm . Let be an effect-valued measure on with values in . Suppose is a bounded measurable function. Consider a sequence of simple functions defined by finite partitions and coefficients , such that . If the sequence converges uniformly to —that is, there exists a sequence of non-negative real numbers such that and for all and —then the sequence of simple integrals converges to the integral in as .
Independence of the integral from the choice of bound
Let be a complete Archimedean order-unit space and be an effect-valued measure on a measurable space with values in . Let be a measurable function. If and are two real numbers such that and for all , then the integral of against defined using the bound is equal to the integral defined using the bound .
The Integral of a Simple Function equals its Simple Integral
Let be a complete Archimedean order-unit space and be an effect-valued measure on a measurable space with values in . Let be a measurable function such that for some real and all . Suppose is a simple function represented by a finite partition of and coefficients such that for all . Then the integral of against coincides with the simple integral: where denotes the indicator function of the set .
Let be a measurable space and be a complete Archimedean order-unit space. Let be an effect-valued measure on with values in . For any measurable function that is bounded by some (i.e., for all ), if is non-negative ( for all ), then its integral against is also non-negative:
Let be a complete Archimedean order-unit space and be an effect-valued measure on a measurable space with values in . For any two measurable functions such that and for all , the integral of their sum against is equal to the sum of their individual integrals: where the integral is defined as the limit of simple functions approximating the bounded measurable functions in the order-unit norm.
Homogeneity of the effect-valued integral:
Let be a complete Archimedean order-unit space and let be an effect-valued measure on a measurable space taking values in . Let be a measurable function such that for all . Then for any scalar , the integral of the scaled function against satisfies the homogeneity property: where the integral is defined as the limit of a sequence of simple functions constructed via uniform mesh approximations.
