Measurable coding of probability measures #
The Giry measurable space on ProbabilityMeasure α is generated by evaluation on measurable
sets, but it is not currently available as a standard Borel space. When α is countably
generated, probability measures nevertheless admit a canonical measurable injective code into a
countable product of ℝ≥0∞: evaluate on the countable set algebra generated by Mathlib's chosen
countableGeneratingSet α.
This code lets probability arguments use a standard Borel target without imposing a topology on
α. Its coordinates generate the Giry measurable space, so every measurable function of a
probability measure into a nonempty standard Borel space factors measurably through the code.
Injectivity follows from uniqueness of finite measures on a generating set algebra.
Main definitions and results #
TauCeti.MeasureTheory.ProbabilityMeasureCodeIndex-- the countable generating set algebra;TauCeti.MeasureTheory.probabilityMeasureCode-- evaluation on every member of that algebra;TauCeti.MeasureTheory.measurable_probabilityMeasureCode-- measurability of the code;TauCeti.MeasureTheory.measurableSpace_probabilityMeasure_eq_comap_probabilityMeasureCode-- the Giry measurable space is induced by the code;Measurable.exists_eq_measurable_comp_probabilityMeasureCode-- measurable factorization through the code;TauCeti.MeasureTheory.probabilityMeasureCode_injective-- the code determines the measure.
The countable index type used to code a probability measure: the set algebra generated by Mathlib's chosen countable generating family for the measurable space.
Equations
Instances For
The measurable evaluation code of a probability measure on a countably generated space.
Equations
Instances For
Evaluation of the probability-measure code.
Every member of the coding set algebra is measurable.
The evaluation code is measurable into its countable product space.
The Giry measurable space on probability measures over a countably generated space is the pullback of the product measurable space along the canonical evaluation code.
A measurable function of a probability measure into a nonempty standard Borel space factors measurably through its canonical code. The extension away from codes of actual probability measures is not specified.
The evaluation code determines a probability measure on a countably generated space.