ECE Ph.D. Qualifying Exam

Communication, Networking, Signal and Image Processing (CS)

Question 1: Probability and Random Processes

August 2015


Solution 1

$ P((Z(t)=0) = P(Z(0)=0, N(t)=Even) + P(Z(0)=1, N(t)=Odd)\\ = pP( N(t)=Even) + (1-p)P( N(t)=Odd)\\ =p\sum_{m=0,1, 2, ...}P(N(t) = 2m)+ (1-p)\sum_{n=0,1,2,...}P(N(t)=2n-1)\\ =p\sum_{m=0,1,2,...}\frac{1}{1+\lambda t}(\frac{\lambda t}{1+\lambda t})^2m + (1-p)\sum_{n=0,1,2,...}\frac{1}{1+\lambda t}(\frac{\lambda t}{1+\lambda t})^{2n-1}\\ =p\cdot\frac{1}{1+\lambda t}\cdot \frac{1}{1-(\frac{\lambda t}{1+\lambda t})^2} + (1-p)\cdot\frac{\lambda t}{1+\lambda t}\cdot\frac{1}{1+\lambda t}\cdot \frac{1}{1-(\frac{\lambda t}{1+\lambda t})^2}\\ =\frac{p+\lambda t}{1+2\lambda t} $

$ P((Z(t)=1) = 1 - P((Z(t)=0) = \frac{1+\lambda t - p}{1+2\lambda t} $

Solution 2

$ P(Z(t)=0)=P(Z(t)=0|N(t)=even)P(N(t)=even)+P(Z(t)=0|N(t)=odd)P(N(t)=odd) $

Note that $ \{Z(t)=0|N(t)=odd\}=\{Z(0)=1\} $ and $ \{Z(t)=0|N(t)=even\}=\{Z(0)=0\} $, therefore,

$ P(Z(t)=0)=P(Z(0)=0)P(N(t)=even)+P(Z(0)=1)P(N(t)=odd)\\ =p\sum_{k=0}^{\infty}\frac{1}{1+\lambda t}\cdot (\frac{\lambda t}{1+\lambda t})^{2k}+(1-p)\sum_{k=0}^{\infty}\frac{1}{1+\lambda t}\cdot (\frac{\lambda t}{1+\lambda t})^{2k+1}\\ =\frac{p}{1+\lambda t}\cdot \frac{1}{1-(\frac{\lambda t}{1+\lambda t})^2}+ \frac{(1-p)\lambda t}{(1+\lambda t)^2}\cdot \frac{1}{1-(\frac{\lambda t}{1+\lambda t})^2}\\ =\frac{p+\lambda t}{1+2\lambda t}\\ P(Z(t)=1) = 1- P(Z(t)=0) = 1-\frac{p+\lambda t}{1+2\lambda t} = \frac{1-p+\lambda t}{1+2\lambda t}\\ P(Z(t)=k)=\left\{ \begin{array}{cc} \frac{p+\lambda t}{1+2\lambda t}, k =0 \\ \frac{1-p+\lambda t}{1+2\lambda t}, k =1\\ 0, else \end{array} \right. $

The solution is correct. However the else case is not necessary. K can only be 0 or 1.

Solution 3

We know that $ Z(t) $ can only take on the values 0 and 1, so we set out to find the probability that $ Z(t) $ = 0; if we subtract this probability from 1, we will have found the probability that $ Z(t) = 1 $, and thus we will have described the entire pmf. We also know that if $ Z(0) $ = 0, $ Z(t) $ must be equal to 0 if $ N(t) $ is even (i.e., if $ k $ is even). Similarly, if $ Z(0)\neq 0, Z(t) $ must be equal to 0 if $ k $ is odd. As such, we can write the expression

$ P(Z(t) = 0) = P(Z(0) = 0, \,N(t)\,\,is\,\,even) + P(Z(0) = 1, \,N(t)\,\,is\,\,odd) = P(Z(0) = 0)P(N(t)\,\,is\,\,even) + P(Z(0) = 1)P(N(t)\,\,is\,\,odd) \\= p\cdot\sum_{i=0}^\infty\frac{1}{1+\lambda t}\left(\frac{\lambda t}{1+\lambda t}\right)^{2i} + (1-p)\cdot\sum_{j=0}^\infty\frac{1}{1+\lambda t}\left(\frac{\lambda t}{1+\lambda t}\right)^{2j+1}. $

We now recall that the sum of an infinite geometric series can be expressed as

$ \sum_{k = 0}^\infty ar^k = \frac{a}{1-r}. $

We can use this to simplify the preceding equation:

$ P(Z(t) = 0) = p\cdot\sum_{i=0}^\infty\frac{1}{1+\lambda t}\left(\frac{\lambda t}{1+\lambda t}\right)^{2i} + (1-p)\cdot\sum_{j=0}^\infty\frac{1}{1+\lambda t}\left(\frac{\lambda t}{1+\lambda t}\right)^{2j+1} \\ = p\cdot\sum_{i=0}^\infty\frac{1}{1+\lambda t}\left(\frac{\lambda t}{1+\lambda t}\right)^{2i} + (1-p)\cdot\frac{\lambda t}{1 + \lambda t}\cdot\sum_{j=0}^\infty\frac{1}{1+\lambda t}\left(\frac{\lambda t}{1+\lambda t}\right)^{2j} \\ = p\cdot\frac{1}{1+\lambda t}\cdot\frac{1}{1-\left(\frac{\lambda t}{1 + \lambda t}\right)^2} + (1-p)\cdot\frac{1}{1+\lambda t}\cdot\frac{\lambda t}{1 + \lambda t}\cdot\frac{1}{1-\left(\frac{\lambda t}{1 + \lambda t}\right)^2}. $

We first combine terms:

$ P(Z(t) = 0) =p\cdot\frac{1}{1+\lambda t}\cdot\frac{1}{1-\left(\frac{\lambda t}{1 + \lambda t}\right)^2} + (1-p)\cdot\frac{1}{1+\lambda t}\cdot\frac{\lambda t}{1 + \lambda t}\cdot\frac{1}{1-\left(\frac{\lambda t}{1 + \lambda t}\right)^2} \\ = \frac{p}{(1+\lambda t)\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)} + \frac{\lambda t}{(1+\lambda t)^2\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)} - \frac{p\lambda t}{(1+\lambda t)^2\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)} \\ = \frac{p + p\lambda t}{(1+\lambda t)^2\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)} + \frac{\lambda t}{(1+\lambda t)^2\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)} - \frac{p\lambda t}{(1+\lambda t)^2\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)} \\ = \frac{p + \lambda t}{(1+\lambda t)^2\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)}. $

Then we simplify the denominator:

$ P(Z(t) = 0) = \frac{p + \lambda t}{(1+\lambda t)^2\left(1-\left(\frac{\lambda t}{1+\lambda t}\right)^2\right)} \\ = \frac{p + \lambda t}{(1+\lambda t)^2 - (\lambda t)^2} \\ = \frac{p + \lambda t}{1 + 2\lambda t}. $

Now that we have found $ P(Z(t) = 0) $, we can easily find $ P(Z(t) = 1) $ by subtracting our result from 1:

$ P(Z(t) = 1) = 1 - P(Z(t) = 0) \\ = 1 - \frac{p + \lambda t}{1 + 2\lambda t}\\ = \frac{1+2\lambda t}{1+ 2\lambda t} - \frac{p + \lambda t}{1 + 2\lambda t} \\ = \frac{1 + \lambda t-p}{1 + 2\lambda t}. $

Similar Problem

Find the mean function $ \mu(t) $ and covariance function $ C_{zz}(t_1,t_2) $ of the process $ Z(t) $.


Back to QE CS question 1, August 2015

Back to ECE Qualifying Exams (QE) page

Alumni Liaison

Ph.D. 2007, working on developing cool imaging technologies for digital cameras, camera phones, and video surveillance cameras.

Buyue Zhang