Scribd
Upload
32 views2 pages

ECE531 Spring 2011 Final Examination

This document provides instructions for a 500 point final exam consisting of 5 problems. Problem 1 involves hypothesis testing with i.i.d. uniform random variables and computing the false positive probability and power function of a proposed detector. Problem 2 involves finding the maximum likelihood estimator for a linear model with uniform noise. Problem 3 involves finding the minimum variance unbiased estimator for an unknown channel impulse response given observations from a known input sequence convolved with the impulse response and Gaussian noise. Problem 4 discusses applying the Rao-Blackwell-Lehmann-Sheffe theorem to an unbiased estimator that is not actually the minimum variance unbiased estimator. Problem 5 involves computing the Kalman filter estimate for a linear dynamical system with Gaussian noise.

Uploaded by

aaymysz
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
32 views2 pages

ECE531 Spring 2011 Final Examination

This document provides instructions for a 500 point final exam consisting of 5 problems. Problem 1 involves hypothesis testing with i.i.d. uniform random variables and computing the false positive probability and power function of a proposed detector. Problem 2 involves finding the maximum likelihood estimator for a linear model with uniform noise. Problem 3 involves finding the minimum variance unbiased estimator for an unknown channel impulse response given observations from a known input sequence convolved with the impulse response and Gaussian noise. Problem 4 discusses applying the Rao-Blackwell-Lehmann-Sheffe theorem to an unbiased estimator that is not actually the minimum variance unbiased estimator. Problem 5 involves computing the Kalman filter estimate for a linear dynamical system with Gaussian noise.

Uploaded by

aaymysz
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

ECE531 Spring 2011 Final Examination

Instructions: This exam is worth a total of 500 points. You may consult two double-sided lettersized sheets of notes (in your own handwriting) and you may use a calculator during the exam.
Please show your work on each problem and box/circle your final answers. Points may be deducted
for a disorderly presentation of your solution. The exam is closed-book.
1. 100 points. Suppose Y0 , Y1 , Y2 , Y3 are i.i.d. observations drawn from U(0, x) where x > 0 is
an unknown deterministic constant. You have two hypotheses:
H0 : x = 1
H1 : x 6= 1.
Someone proposes to use the following detector:
(
1 max Yi < 21 or max Yi > 1
(Y0 , Y1 , Y2 , Y3 ) =
0 otherwise.
(a) (40 points) Compute the false positive probability of this decision rule.
(b) (60 points) Compute and sketch the power function (x) of this decision rule.
Hint: Let Z = max Yi . The pdf of Z parameterized by x is
( 3
4z
0zx
4
px (z) = pZ (z ; x) = x
0
otherwise.
2. 100 points. Suppose we observe the sequence Y0 , Y1 , . . . , Yn1 given by
Yk = + Uk
i.i.d.

for k = 0, . . . , n 1 where Uk U(0.5, 0.5) and R is an unknown non-random


parameter.
Pn1

yk may not be a good


(a) (25 points) Explain why the sample mean (y)
= n1 k=0
estimator of in this problem. Hint: Consider an example with three observations.
(b) (50 points) Find a/the maximum-likelihood estimator (MLE) of .
(c) (25 points) Is your MLE unique? Explain.
3. 100 points. Consider the scenario shown in Figure 1 with the following notation:
The known input sequence is s0 , s1 , . . . .
The unknown non-random channel impulse response is = [1 , 2 ] R2 .
1

channel with
unknown
impulse response

known
input
sequence

observations
zero-mean i.i.d.
Gaussian noise

Figure 1: Channel estimation problem scenario.


i.i.d.

The Gaussian noise is W0 , W1 , . . . with Wk N (0, 1).


The observations are Y0 , Y1 , . . . .
Suppose [s0 , s1 , s2 ] = [1, 2, 1], i.e. a known input sequence of three numbers was sent through
the channel. Note that there will be four observations Y0 , Y1 , Y2 , Y3 in this case because the
output of the channel is simply the convolution of the input sequence with the unknown
non-random two-sample channel impulse response. Find the MVUE of the unknown channel
impulse response .
i.i.d.

4. 100 points. Suppose you observe samples Yk N (, 1) for k = 0, 1, . . . , n 1 where


R is a scalar unknown random parameter. A sufficient statistic for is, of course,
T (Y ) = [Y0 , . . . , Yn1 ] since the whole observation is always sufficient. Suppose someone
) = Y0 as any old unbiased estimator and to apply the Rao-Blackwellsuggests to use (Y
Lehmann-Sheffe theorem to compute the MVUE by evaluating the conditional expectation
n
o
) | T (Y ) = T (y) .
MVU? (y) = E (Y
(1)
) = Y0 is an unbiased estimator of .
(a) (25 points) Confirm (Y
(b) (25 points) Compute MVU? (y) in (1).
(c) (50 points) Explain why MVU? (y) in (1) is not the MVUE of . Why doesnt the RBLS
theorem give the MVUE in this case?

5. 100 points. Consider the scalar dynamical system


X[ + 1] = X[] + U [] for = 0, 1, . . .
with observations given by Y [] = X[] + V [] for = 0, 1, . . . . Assume that {U [0], U [1], . . . }
and {V [0], V [1], . . . } are independent sequences of i.i.d. Gaussian random variables with
i.i.d.

i.i.d.

U [k] N (0, 1) and V [k] N (0, 1). Also assume that the initial state X[0] N (0, 1) and
is independent of all U [] and V [].
(a) (50 points) Given Y [0] = 1 and Y [1] = 2 compute the Kalman filter state estimate
| 1], i.e. the Kalman filter estimate of the state X[1] given the observations Y =
X[1
[Y [0], Y [1]] .
| 1] is the MMSE estimate of the state X[1] given the obser(b) (50 points) Confirm that X[1

vations Y = [Y [0], Y [1]] by directly computing the MMSE estimate via the conditional
expectation, i.e. compute
mmse = E{X[1] | Y [0], Y [1]}.
X

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy