Skip to content

Appendix 5: Function Spaces and Measure Theory

Mathematical Preliminaries for Continuous Ensembles

This appendix provides the rigorous measure-theoretic foundations used in Parts III–V.


Measurable Spaces

A measurable space \((\Omega, \mathcal{F})\) consists of:
- \(\Omega\): sample space (set of all possible outcomes)
- \(\mathcal{F}\): \(\sigma\)-algebra (collection of measurable subsets)

For function spaces, \(\Omega\) is typically a space of functions \(f: [0,T] \to \mathbb{R}\).


Probability Measure

A probability measure \(P: \mathcal{F} \to [0,1]\) satisfies:
1. \(P(\Omega) = 1\)
2. Countable additivity: \(P(\bigcup_i A_i) = \sum_i P(A_i)\) for disjoint \(A_i\)


Kolmogorov Extension Theorem

Given consistent finite-dimensional distributions, there exists a unique probability measure on the infinite-dimensional space.

Consistent means: If you marginalize the joint distribution of \((f(t_1), \ldots, f(t_n))\) to any subset of time points, you get the corresponding lower-dimensional distribution.


Important Function Spaces

\(L^2\) Space

\[L^2[0,T] = \left\{f: \int_0^T |f(t)|^2 \, dt < \infty\right\}\]

Hilbert space with inner product \(\langle f, g \rangle = \int f(t) g(t) \, dt\).

Most physical signals belong to \(L^2\) (finite energy).

\(L^\infty\) Space

\[L^\infty[0,T] = \left\{f: \text{ess sup}_{t \in [0,T]} |f(t)| < \infty\right\}\]

Bounded functions. Important for peak power constraints.

Sobolev Spaces

Functions with derivatives in \(L^2\). Used when smoothness matters.


Stochastic Processes as Random Functions

A stochastic process \(\{X_t\}_{t \in T}\) is a collection of random variables indexed by time. Equivalently, it is a single random function:

\[X: \Omega \to \mathbb{R}^T\]

where \(\mathbb{R}^T\) is the space of all functions \(T \to \mathbb{R}\).

The law of the process is the probability measure induced on \(\mathbb{R}^T\).


Stationarity and Ergodicity

Strict Stationarity

For all \(n\), all \(t_1, \ldots, t_n\), all \(\tau\):

\[(X_{t_1}, \ldots, X_{t_n}) \stackrel{d}{=} (X_{t_1+\tau}, \ldots, X_{t_n+\tau})\]

Wide-Sense Stationarity

Weaker: \(\mathbb{E}[X_t] = \mu\) (constant) and \(\mathbb{E}[X_t X_{t+\tau}] = R(\tau)\) (depends only on lag).

Ergodicity

Time averages = ensemble averages (almost surely).

For Gaussian processes: ergodicity can be checked from the spectral density.


Wiener Measure

The Wiener measure is the probability measure on continuous functions corresponding to Brownian motion:

  • \(W(0) = 0\)
  • Independent increments
  • \(W(t) - W(s) \sim \mathcal{N}(0, t-s)\)

This measure is concentrated on nowhere-differentiable functions — almost all Brownian paths are rough!


Application to Information Theory

For continuous-time Gaussian channels, the rigorous formulation uses:
- Cameron-Martin space (functions in the support of the Wiener measure)
- Radon-Nikodym derivative (likelihood ratio)
- Mutual information defined via Kullback-Leibler divergence of path measures


Next: Appendix 6 — Continuous Entropy Properties