Stochastic Fubini theorem (source code)

= Stochastic Fubini theorem
{title2=$\int du\int H(s,u)dW_s=\int(\int H(s,u)du)dW_s$}

A stochastic Fubini theorem permits interchange of a parameter <integral> with a <stochastic integral>. One sufficient version uses a <measure space> $(U,\mu)$ with $\mu(U)<\infty$ and a function $H(s,\omega,u)$ measurable for the product of the <predictable sigma-algebra> and the parameter <sigma-algebra> and satisfies
$$
\int_U\mathbb E\int_0^T|H(s,u)|^2ds\,\mu(du)<\infty.
$$
Then both sides below exist in <L2 space> and are equal:
$$
\int_U\left(\int_0^TH(s,u)\,dW_s\right)\mu(du)
=\int_0^T\left(\int_UH(s,u)\mu(du)\right)dW_s.
$$
For elementary parameter functions the identity is linearity of the <Itô integral>. Approximation in the displayed squared norm, the <Itô isometry> and the <Cauchy-Schwarz inequality> pass to the limit. Finite parameter measure is part of this sufficient version; more general versions require different integrability bounds. For example, inserting the deterministic indicator $\mathbf1_{\{s\leq u\}}$ on $[0,T]^2$ gives $\int_0^TW_u\,du=\int_0^T(T-s)\,dW_s$.