27.2 Exercises
Exercise 27.1 (Inspired by Logan and Wolesensky (2009)) Let \(R(t)\) denote the rainfall at a location at time \(t\), which is a random process. Assume that probability of the change in rainfall from day \(t\) to day \(t+\Delta t\) is the following:
change | probability |
---|---|
\(\Delta R = \rho\) | \(\lambda \Delta t\) |
\(\Delta R = 0\) | \(1- \lambda \Delta t\) |
The stochastic differential equation generated by this process is \(dR = \lambda \rho \; dt + \sqrt{\lambda \rho^{2}} \; dW(t)\).
- What is the Fokker-Planck partial differential equation for the probability distribution \(f(R,t)\)?
- What is a formula that solves the Fokker-Planck partial differential equation?
- Make some representative plots of the solution as it evolves over time
Exercise 27.2 A particle is moving in a gravitational field but still allowed to diffuse randomly. In this case the stochastic differential equation is \(dx = -g \; dt + \sqrt{D} \; dW(t)\).
- What is the Fokker-Planck partial differential equation for the probability distribution \(f(x,t)\)?
- Based on the work done in this section, what is the equation for the probability distribution \(f(x,t)\)?
Exercise 27.3 Consider the stochastic differential equation \(\displaystyle dS = \left( 1 - S \right) + \sigma dW(t)\), where \(\sigma\) controls the amount of stochastic noise.
- First let \(\sigma = 0\) so the equation is entirely deterministic. Classify the stability of the equilibrium solutions for this differential equation.
- Still let \(\sigma = 0\). Apply separation of variables to solve this differential equation.
- Now let \(\sigma = 0.1\). Do 100 realizations of this stochastic process, with initial condition \(S(0)=0.5\). What do you notice?
- Now try different values of \(\sigma\) larger and smaller than 0.1. What do you notice?
- What is the Fokker-Planck partial differential equation for the probability distribution \(f(S,t)\)?
Exercise 27.4 (Inspired by Logan and Wolesensky (2009)) Consider the differential equation \(\displaystyle x' = \lambda x - c \mu x^{2}\), which is similar to a logistic differential equation. The per capita rate equation for this differential equation is \(\displaystyle \frac{x'}{x} = \lambda - c \mu x\).
- Assume there is noise to this per capita rate, i.e. \(\displaystyle \frac{x'}{x} \rightarrow \displaystyle \frac{x'}{x} + \mbox{ Noise}\). With this revised equation, what are the deterministic and stochastic parts?
- What is the Fokker-Planck partial differential equation for the probability distribution \(f(x,t)\)?
- The assumes that \(f_{t} = 0\), so \(f(x,t) \rightarrow f(x)\). Through direct verification, show that \(f(x)=Dx^{\alpha-1} e^{-\beta x}\) is a solution to the steady-state distribution, where \(\alpha = 2 \lambda - 1\) and \(\beta = 2c\mu\).
- With \(\lambda = c = \mu = D = 1\), make a graph of \(f(x)\) and graph it below:
Exercise 27.5 (Inspired by Gardiner (2004)) A type of chemical reaction is \(X + A \leftrightarrow 2X\), where \(A\) acts like an enzyme. The stochastic differential equation that describes this scenario is:
\[\begin{equation} dX = \left( A X - X^{2} \right) \; dt + \left( AX + X^{2} \right) dW(t) \end{equation}\]
- What is the Fokker-Planck equation for this stochastic differential equation?
- The steady-state distribution for this process is \(p(X) = e^{-2X}(A+X)^{4A-1}X^{-1}\). With \(A=1\) make a plot of this distribution.
Exercise 27.6 Models of cell membranes take account for the energy needed for ions and other materials to cross the cell membrane, usually expressed as a membrane potential \(U(x)\), where \(x\) is the current position of a particle distance. The probability \(p\) of the particle being at position \(x\) at time \(t\) is given by the Fokker-Planck equation:
\[\begin{equation} v \frac{\partial p}{\partial t} = \frac{\partial}{\partial x} \left( U'(x) p \right) + k T \frac{\partial^{2} p}{\partial x^{2}}, \end{equation}\]
where \(k\) is Boltzmann’s constant and \(T\) is the temperature.
- Write the Fokker-Planck equation in steady state. Show that a solution for the steady-state equation is \(p(x)=C e^{-\frac{U(x)}{kT}}\)
- If the steady-state distribution is normal, what is function will \(U(x)\) be?