stochastic differential equation: Nonlinear Function
Created: August 27, 2022
Modified: August 29, 2022

stochastic differential equation

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 It might be incomplete, wrong, outdated, or stupid.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.

SDEs are typically written in terms of the differential dWdW of a Weiner process (Brownian motion), e.g.,

dx=αt(x)dt+βt(x)dWdx = \alpha_t(x)dt + \beta_t(x)dW

Although Weiner processes are nowhere differential, this notation is given meaning by the Itô integral which describes a procedure to evaluate the relevant integrals.

Background: an ordinary differential equation

dxdt=at(x)\frac{dx}{dt} = a_t(x)

can be equivalently written in differential form

dx=at(x)dtdx = a_t(x)dt

and also integral form

xs=x0+t0sat(xt)dt.x_s = x_0 + \int_{t_0}^s a_t(x_t) dt.

Similarly a stochastic differential equation may be specified as a differential ODE on realizations of Gaussian white noise ξt(ω)N(0,1)\xi_t(\omega) \sim \mathcal{N}(0, 1):

dxt(ω)=αt(xt(ω))dt+βt(xt(ω))ξt(ω)dtdx_t(\omega) = \alpha_t(x_t(\omega))dt + \beta_t(x_t(\omega))\xi_t(\omega)dt

or in the equivalent integral form:

xs(ω)=x0(ω)+t0sαt(xt(ω))dt+t0sβt(xt(ω))ξt(ω)dt.x_s(\omega) = x_0(\omega) + \int_{t_0}^s \alpha_t(x_t(\omega)) dt + \int_{t_0}^s \beta_t(x_t(\omega)) \xi_t(\omega)dt.

Appealing to the Itô integral formalism we recast the second integral in terms of the differential of a Wiener process ξt(ω)dt=dWt(ω)\xi_t(\omega)dt = dW_t(\omega), which provides formal justification for writing the SDE in the simpler form

dx=αt(x)dt+βt(x)dW,dx = \alpha_t(x)dt + \beta_t(x)dW,

where we view x(t)x(t) as being 'driven' by an underlying Weiner process.