Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
129 views

Homework 2

This document contains 6 homework problems related to linear system theory: 1. Show that the set of points closer to point a than b forms a halfspace and provide the explicit form. 2. Determine whether several statements about matrix ranks and nullspaces are true or false. 3. Show properties about fundamental subspaces and determinants of matrices. 4. Derive properties of the characteristic polynomial of a square matrix. 5. Use Schur's theorem to prove Cayley-Hamilton theorem about characteristic polynomials of matrices. 6. Use eigenvalues and eigenvectors to detect communities in a social network from its connectivity matrix.

Uploaded by

jerry
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
129 views

Homework 2

This document contains 6 homework problems related to linear system theory: 1. Show that the set of points closer to point a than b forms a halfspace and provide the explicit form. 2. Determine whether several statements about matrix ranks and nullspaces are true or false. 3. Show properties about fundamental subspaces and determinants of matrices. 4. Derive properties of the characteristic polynomial of a square matrix. 5. Use Schur's theorem to prove Cayley-Hamilton theorem about characteristic polynomials of matrices. 6. Use eigenvalues and eigenvectors to detect communities in a social network from its connectivity matrix.

Uploaded by

jerry
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

ELEC505/ELEC405 Handout #3

Linear System Theory Sunday, November 8, 2020

Homework Set #2
Due: Sunday, November 22 , 2020.

1. Nearest Neighbor partitioning and halfspaces...


Suppose a, b ∈ Rn are two given points. Show that the set of points in Rn that are
closer to a than b is a half space, i.e.:

{x |k x − a k≤k x − b k} = {x | cT x ≤ d}

for appropriate c ∈ Rn and d ∈ R. Provide c and d explicitly (in terms of a and b),
and draw a picture showing a, b, c and the halfspace (for n=2, hand-drawing is fine
let’s give Matlab/Python/Julia stuff a break... you pick your favorite a, b.). Remember
that the set {x | cT x = d} (border) is called an hyperplane.

2. Rank and Matrix Products....


For each of the following statements, either show that it is true or give a counterexample

(a) If AB is full rank then A and B are full rank.


(b) If A and B are full rank then AB is full rank.
(c) If A and B have zero nullspace, then so does AB.
(d) If A and B are onto, then so is AB.

3. Fundamental Subspaces and Determinant....


Given A ∈ Rm×n show that

(a) If R(A) = Rm , then AAT is full rank.


(b) If N (A) = {0} then AT A is full rank.

4. Characteristic Polynomial of a Square Matrix....


Consider the characteristic polynomial X(s) = det(sI − A) of the matrix A ∈ Rn×n

(a) Show that X(s) is monic, which means that its leading coefficient is one: X(s) =
sn + ...
(b) Show that sn−1 coefficient of X(s) is given by − Tr A. (Remember Tr A is the
trace of the matrix A which is the sum of the diagonal entries of A, i.e., Tr A =
Pn
i=1 Aii .)

(c) Show that the constant coefficient of X(s) is given by det(−A).

1
(d) Let λ1 , ..., λn denote the eigenvalues of A, so that

X(s) = sn + an−1 sn−1 + ... + a1 s + a0 = (s − λ1 )(s − λ2 )...(s − λn )

by equating coefficients show that an−1 = − ni=1 λi and a0 =


P Qn
i=1 (−λi ). There-
fore conclude that Tr A = ni=1 λi and detA = ni=1 λi
P Q

5. Schur’s Theorem and Cayley-Hamilton Theorem...


Schur’s Theorem is as follows: Given A ∈ Rn×n with eigenvalues λ1 , ..., λn in any
prescribed order, there is a unitary matrix U such that

U∗ AU = T

is upper triangular with diagonal entries tii = λi , i = 1, ..., n. Remember that a unitary
matrix U has the property U∗ U = UU∗ = I where U∗ is the hermitian transpose of
U.

(a) (Cayley-Hamilton Theorem) Using Schur’s Theorem above, prove the Cayley-
Hamilton Theorem which states: Let pA (t) be the characteristic polynomial of
A ∈ Rn×n , (i.e., pA (t) = det(tI − A)), then

pA (A) = 0

Note: Don’t just plug in A into determinant expression. First right the nth
characteristic polynomial in terms of t’s powers and then plug in A to replace t.
In other words, you should write a polynomial containing powers of A and show
that this polynomial is equal to the zero matrix.
(b) Using the result of (a) and assuming

pA (t) = tn + an−1 tn−1 + ... + a1 t + a0

write each of An , An+1 , An+2 as polynomials of A with degree at most n − 1.


Assuming A is invertible write A−1 also as a polynomial of A of degree at most
n − 1.

2
6. Theory in Action for Big Data!: Eigenvectors as community detectors

Figure 1: Two team fan communities and their facebook connectivity.

Now consider a population of 2N people,. Half of the population is part of community


A, the other half is part of community B. All these guys have facebook accounts where
they are potentially connected as illustrated in Figure 1. Two people in the same
community are connected with probablility p and two people in different communities
are connected with probability q < p. Connections are probabilistically independent.
We define the facebook connectivity matrix C as
(
1 if i, j connected,
Cij =
0 otherwise

It is a bit weird but we will assume that to be true for Cii too... Anyway, we define
IA ⊂ {1, . . . , 2N } as the indices of the people in community A, and IB = I¯A .

(a) Assume that IA = {1, . . . , N } and IB = {N + 1, . . . , 2N }. Find the matrix E(C),


where E(·) is the expected value operator.
(b) What is the rank of E(C) in the previous part?
(c) What are the eigenvalues and eigenvectors of E(C) in part a? How would it
change if IA is arbitrary? How would you tell communities, i.e., IA and IB from
eigenvalues, eigenvectors?
(d) Now going back to real life. We don’t have p, q and E(C) available. We have
only C (It appears we can assume Facebook is willing to sell this information!).
As a relief, the members of our geeky theoretical data science team claim (and
prove) that for a relatively large population, eigenvectors of C, corresponding to
top eigenvalues, are fairly close to eigenvectors of E(C). Using this information,
here is the task:
• Take N = 100. Randomly generate IA and C. (Take p = 0.6, q = 0.1).

3
• Display C using imshow and print. Can you tell communities from this
picture?
• Check out eigenvalues, eigenvectors? Are they helpful in detecting IA ? How?
Display relevant eigenvector(s) using stem function.
• How accurate is your result? (Run this test 1000 times and report the average
percentage of correct identification of each person’s community.)

Please include your plots as printouts attached to your handwritten solution with clear
labeling including problem number.

You might also like