Math 510 Final Exam: Invariant Factors, Matrices, SVD, and Hermitian, Exams of Linear Algebra

The final examination for math 510, covering topics such as determining invariant factors, normal matrices, idempotent matrices, singular value decomposition, and hermitian matrices. The examination includes questions on finding the rational canonical form, proving that the adjoint of a normal matrix commutes with any commuting matrix, proving that the rank of an idempotent matrix equals the number of nonzero eigenvalues, finding the reduced singular value decomposition, and proving that the schur complement of a hermitian matrix is hermitian.

Typology: Exams

Pre 2010

Uploaded on 09/02/2009

koofers-user-p3f
koofers-user-p3f 🇺🇸

10 documents

1 / 3

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Math 510 Final Examination 12 December 2006
Directions: All answers must be justified by computation or explanation. Greater weight will be given to one
whole (correct) solution than to two error-free but incomplete solutions. Six complete correct answers will receive
full credit, but you may answer additional questions if desired. Write each solution on a separate page. Submit
solutions in the same order as the questions.
1. The characteristic polynomial of AQ11×11 is p(x) = (x1)5(x2+ 1)3, the minimum polynomial
of Ais p(x) = (x1)3(x2+ 1)2, and the geometric multiplicity of eigenvalue λ= 1 is 3. Determine
the invariant factors of A.
2. Let Abe a normal matrix. Prove that Acommutes with any matrix that commutes with A.
3. (a) Assume AFn×nis idempotent. Prove that the rank of Ais equal to the number of nonzero
eigenvalues of A.
(b) Give a counterexample to show that the statement in 3a may be false without the assumption
that Ais idempotent.
4. Find the reduced singular value decomposition (i.e., find singular values and right and left singular
vectors for nonzero singular values) for A=
1 2
2 0
0 2
.
5. Let A, B Cn×nbe Hermitian. Prove that λmax(AB) + λmin(B)λmax (A).
6. Let ACn×nand λC. Prove that
rank((AλI)2) + rank((AλI )8)rank((AλI )5) + rank((AλI)5).
7. Let AFn×nand let λσ(A). Prove that Ais similar to a matrix having the sum of the elements
in each row equal to λ.
8. Let HCn×nbe Hermitian and partition Has H=H11 H12
H21 H22with H11 , H22 square and H11
nonsingular. Prove that the Schur complement H/H11 is Hermitian.
1
pf3

Partial preview of the text

Download Math 510 Final Exam: Invariant Factors, Matrices, SVD, and Hermitian and more Exams Linear Algebra in PDF only on Docsity!

Math 510 Final Examination 12 December 2006

Directions: All answers must be justified by computation or explanation. Greater weight will be given to one whole (correct) solution than to two error-free but incomplete solutions. Six complete correct answers will receive full credit, but you may answer additional questions if desired. Write each solution on a separate page. Submit solutions in the same order as the questions.

  1. The characteristic polynomial of A ∈ Q^11 ×^11 is p(x) = (x − 1)^5 (x^2 + 1)^3 , the minimum polynomial of A is p(x) = (x − 1)^3 (x^2 + 1)^2 , and the geometric multiplicity of eigenvalue λ = 1 is 3. Determine the invariant factors of A.
  2. Let A be a normal matrix. Prove that A∗^ commutes with any matrix that commutes with A.
  3. (a) Assume A ∈ F n×n^ is idempotent. Prove that the rank of A is equal to the number of nonzero eigenvalues of A. (b) Give a counterexample to show that the statement in 3a may be false without the assumption that A is idempotent.
  4. Find the reduced singular value decomposition (i.e., find singular values and right and left singular

vectors for nonzero singular values) for A =

  1. Let A, B ∈ Cn×n^ be Hermitian. Prove that λmax(A − B) + λmin(B) ≤ λmax(A).
  2. Let A ∈ Cn×n^ and λ ∈ C. Prove that

rank((A − λI)^2 ) + rank((A − λI)^8 ) ≥ rank((A − λI)^5 ) + rank((A − λI)^5 ).

  1. Let A ∈ F n×n^ and let λ ∈ σ(A). Prove that A is similar to a matrix having the sum of the elements in each row equal to λ.
  2. Let H ∈ Cn×n^ be Hermitian and partition H as H =

[

H 11 H 12

H 21 H 22

]

with H 11 , H 22 square and H 11 nonsingular. Prove that the Schur complement H/H 11 is Hermitian.

  1. The characteristic polynomial of A ∈ Q^11 ×^11 is p(x) = (x − 1)^5 (x^2 + 1)^3 , the minimum polynomial of A is p(x) = (x − 1)^3 (x^2 + 1)^2 , and the geometric multiplicity of eigenvalue λ = 1 is 3. Determine the invariant factors version of the rational canonical form of A, i.e., determine RCFIF (A). Find Jordan form first: Eigenvalue 1: Geometric multiplicity is 3, so 3 Jordan blocks of size n 1 , n 2 , n 3. Largest block size n 1 = 3. Algebraic multiplicity is 5, i.e., 3 + n 2 + n 3 = 5, i.e., n 2 = n 3 = 1. Thus for λ = 1, J 3 (1) ⊕ J 1 (1) ⊕ J 1 (1) and the elementary divisors are (x − 1)^3 , x − 1 , x − 1. Eigenvalue ±i: Largest block size n 1 = 2. Algebraic multiplicity is 3, i.e., 2 + n 2 = 3, i.e., n 2 = 1. Thus for λ = i (−i is analogous), J 2 (i) ⊕ J 1 (i). For ±i together, the elementary divisors are (x^2 + 1)^2 , x^2 + 1. The invariant factors are: (x − 1)^3 (x^2 + 1)^2 , (x − 1)(x^2 + 1), x − 1.
  2. Let A be a normal matrix. Prove that A∗^ commutes with any matrix that commutes with A.

It was shown in Theorem 8.1 that A normal implies there exists a polynomial f (x) such that A∗^ = f (A). If AB = BA, then A∗B = f (A)B = Bf (A) = BA∗.

  1. (a) Assume A ∈ F n×n^ is idempotent. Prove that the rank of A is equal to the number of nonzero eigenvalues of A. Any idempotent matrix is similar to a diagonal matrix with 0s and 1s on the diagonal (because its minimal polynomial divides x(x − 1)). The rank of such a diagonal matrix is the number of nonzero eigenvalues, and both rank and spectrum are invariant under similarity. (b) Give a counterexample to show that the statement in 3a is false without the assumption that A is idempotent.

A =

[

]

  1. Find the reduced singular value decomposition (i.e., find singular values and right and left singular

vectors for nonzero singular values) for A =

A∗A =

[

]

. pA(x) = x^2 − 13 x + 36 = (x − 9)(x − 4), so σ 1 = 3, σ 2 = 2 are the singular values.

Find right singular vectors: A∗A − 9 I =

[

]

, so v 1 = √^15

[

]

A∗A − 4 I =

[

]

, so v 2 = √^15

[

]

Find left singular vectors (for σ 1 , σ 2 ): u 1 = (^) σ^11 Av 1 = 3 √^15

u 2 = (^) σ^12 Av 2 = √^15

  1. Let A, B ∈ Cn×n. Prove that λmax(A − B) + λmin(B) ≤ λmax(A).

It is a consequence of Rayleigh-Ritz (also Theorem 7.10 in Zhang) that

λmax(C + G) ≤ λmax(C) + λmax(G).