Itô integral: Nonlinear Function
Created: August 27, 2022
Modified: August 30, 2022

Itô integral

This page is from my personal notes, and has not been specifically reviewed for public consumption. It might be incomplete, wrong, outdated, or stupid. Caveat lector.

This is the technical formulation that makes it meaningful to write stochastic differential equations 'driven by' a Weiner process WW, such as

dx=α(x,t)dt+β(x,t)dW.dx = \alpha(x, t) dt + \beta(x, t)dW.

The construction is nontrivial because the Wiener process is almost surely nowhere differentiable, so the differential dWdW is not defined in the conventional sense. The Itô integral is essentially a procedure for evaluating integrals with respect to dWdW and other stochastic processes, which (following the fundamental theorem of calculus) defines the corresponding differentials. Essentially, we

The full construction is built up from the basic case of integrating a constant function β(x,t;ω)=βˉ\beta(x, t; \omega) = \bar{\beta} over a range: for any sample path W(ω)W(\omega) we define

I[βˉ](ω)s=t0tβˉdWs(ω)=βˉs=t0tdWs(ω)βˉ(Wt(ω)Wt0(ω))\begin{align*} I[\bar\beta](\omega) &\coloneqq \int_{s=t_0}^t \bar{\beta} dW_s(\omega)= \bar\beta \int_{s=t_0}^t dW_s(\omega)\\ &\coloneqq \bar{\beta} \left(W_t(\omega) - W_{t_0}(\omega)\right) \end{align*}

This can be generalized by appropriate constructions to define integrals I[β](ω)I[\beta](\omega) of piecewise constant functions β(t)\beta(t), random piecewise constant functions β(t;ω)\beta(t; \omega), and arbitrary random functions β(t;ω)\beta(t; \omega) built up as limits of piecewise constant functions. Note that the integrand β\beta must be adapted to the same filtration as WtW_t; i.e., it can't depend on future information, only what would be available at time tt.