0
$\begingroup$

The algorthim use for updating the voltage of each componant at every time step is $$v(t+\Delta t)=\begin{cases} v(t)e^{-\gamma \Delta t},& n=0\\ v(t)e^{-\gamma \Delta t}+h_{*}& n=1\end{cases}$$ where $n$ denotes the number of impulses arriving at a componant in the interval $(t,t+\Delta t).$ Here $ h_{*}=0.03, \gamma=20, \Delta t=10^{-5}.$ I have no idea how to use this algorithm in monte carlo simulation. More precisely , I have no idea about the use of monte carlo simulation. Plz help

$\endgroup$
2
$\begingroup$

It seems to me you would just be randomly selecting $n$ at each time step, given some distribution you should be sampling $n$ from, and using your dynamics function to propagate the voltage from time $t$ to time $t + \Delta t$.

You would then run multiple Monte Carlo simulations from the initial time to the final time where you basically select a random value for $n$ at each time step. Using the final results, you could compute statistical information like the expected value for the voltage at some time $\hat{t}$, the variance of your voltage at time $\hat{t}$, etc.

If you write a code for this and use the values for the constants the way you set them, you might produce the following type of graphic:

enter image description here

$\endgroup$
  • 1
    $\begingroup$ it's also worth considering which distribution $n$ should be drawn from at every time step. a natural choice would be to flip a fair coin - Bernoulli with $p=0.5$, so previous draws are "forgotten". other choices are also possible, depending on the behavior you're modeling. $\endgroup$ – GoHokies Jun 3 '17 at 9:52
  • $\begingroup$ @GoHokies Good point to mention, I will update my answer! $\endgroup$ – spektr Jun 3 '17 at 15:09
  • $\begingroup$ @C.Howard in this experiment the distribution is poisson. $\endgroup$ – jixon Jun 4 '17 at 15:33
0
$\begingroup$

If $\Delta t$ is small enough then $n$ can be only 0 or 1. Otherwise you must either reduce $\Delta t$ or change your formula to $v(t+\Delta t) = v(t)e^{-\gamma \Delta t} + n \times h_{*}$ (which by the way also works for $n=0$).

And if $n$ can be only 0 or 1, there is no notion of distribution: only of probability. The probability of $n$ being 1 is $p_0 \times \Delta t$ (to first order, so $\Delta t$ must be small).

Then it is just a matter of drawing a (pseudo)random number between 0 and 1 (equiprobably): if it is smaller than $p_0 \times \Delta t$ then $n:=1$, otherwise $n:=0$. The probability of a random number between 0 and 1 being less than $p_0 \times \Delta t$ is... $p_0 \times \Delta t$, which is exactly the probability you want to set $n$ to 1.

$\endgroup$
  • $\begingroup$ if $n$ can be only 0 or 1, there is no notion of distribution: only of probability - are you sure that is correct? in fact, your answer does define a particular type of random distribution for $n$. also, what is $p_0$? $\endgroup$ – GoHokies Jun 4 '17 at 10:06
  • $\begingroup$ My point is that distribution information boils down to a single number: a probability. $p_0$ is the number of events per unit time, i.e. $1/p_0$ is the frequency of events. $\endgroup$ – Mathieu Bouville Jun 5 '17 at 9:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.