nLab filtered probability space

Contents

Context

Measure and probability theory

Limits and colimits

Contents

Idea

In probability theory, a sigma-algebra can be interpreted as a system of possible “distinctions” that can be made in a measurable or probability space XX.

A filtration on a probability space is an increasing system of sigma-algebras, for example indexed by time, which one can interpret as “being able to make more and more distinctions”, or “learning more and more as time progresses”.

As the intuition may suggest, it is an instantiation of the categorical notion of filtration, in the context of probability theory.

Definition

A filtered probability space consists of a probability space (X,𝒜,p)(X,\mathcal{A},p), equipped with a filtration of sub-sigma-algebras of 𝒜\mathcal{A}, i.e. a collection ( i) iI(\mathcal{F}_i)_{i\in I} where:

  • II is a partially ordered set (often totally ordered), with finitary upper bounds;
  • For each iIi\in I, i\mathcal{F}_i is a sub-sigma-algebra of 𝒜\mathcal{A};
  • For each ijIi\le j\in I, i j\mathcal{F}_i\subseteq \mathcal{F}_j.

Note that a filtration ( i) iI(\mathcal{F}_i)_{i\in I} as defined above is a filtered object in the category of sigma-algebras, but a cofiltered object? in the category of measurable or probability spaces. Indeed, if i j\mathcal{F}_i\subseteq \mathcal{F}_j, the set-theoretic identity map XXX\to X is measurable as a function (X, j)(X, i)(X,\mathcal{F}_j)\to(X,\mathcal{F}_i), but not the other way around.

Limits and colimits

(…)

filtered objects

associated graded objects

References

  • Ruben Van Belle, A categorical treatment of the Radon-Nikodym theorem and martingales, 2023. (arXiv)

  • Paolo Perrone and Ruben Van Belle, Convergence of martingales via enriched dagger categories, 2024. (arXiv)

Introductory material

category: probability

Last revised on July 21, 2024 at 19:38:30. See the history of this page for a list of all contributions to it.