Scribd Logo
Upload

Welcome to Scribd!

  • 9 views

    Exam 3

    Uploaded by

    ashok kumar
    This document outlines 6 computational linear algebra exam problems: 1) Prove that the range of L1 is contained in the range of L2 if and only if there exists L0 such that L1 = L2 o L0. 2) Apply Gram-Schmidt decomposition to show that an invertible matrix A can be decomposed as A = QR, where Q is orthogonal and R is upper triangular. Explain how to solve linear systems and compute determinants using this decomposition. 3) Prove that a matrix P is an orthogonal projector if and only if P2 = P and P^T = P. 4) Prove that if A and B are symmetric positive

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Exam 3

    Uploaded by

    ashok kumar
    9 views2 pages
    This document outlines 6 computational linear algebra exam problems: 1) Prove that the range of L1 is contained in the range of L2 if and only if there exists L0 such that L1 = L2 o L0. 2) Apply Gram-Schmidt decomposition to show that an invertible matrix A can be decomposed as A = QR, where Q is orthogonal and R is upper triangular. Explain how to solve linear systems and compute determinants using this decomposition. 3) Prove that a matrix P is an orthogonal projector if and only if P2 = P and P^T = P. 4) Prove that if A and B are symmetric positive

    Original Description:

    hvv

    Original Title

    exam 3 (1)

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document outlines 6 computational linear algebra exam problems: 1) Prove that the range of L1 is contained in the range of L2 if and only if there exists L0 such that L1 = L2 o L0. 2) Apply Gram-Schmidt decomposition to show that an invertible matrix A can be decomposed as A = QR, where Q is orthogonal and R is upper triangular. Explain how to solve linear systems and compute determinants using this decomposition. 3) Prove that a matrix P is an orthogonal projector if and only if P2 = P and P^T = P. 4) Prove that if A and B are symmetric positive

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    9 views2 pages

    Exam 3

    Uploaded by

    ashok kumar
    This document outlines 6 computational linear algebra exam problems: 1) Prove that the range of L1 is contained in the range of L2 if and only if there exists L0 such that L1 = L2 o L0. 2) Apply Gram-Schmidt decomposition to show that an invertible matrix A can be decomposed as A = QR, where Q is orthogonal and R is upper triangular. Explain how to solve linear systems and compute determinants using this decomposition. 3) Prove that a matrix P is an orthogonal projector if and only if P2 = P and P^T = P. 4) Prove that if A and B are symmetric positive

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 2

    Exam – Computational Linear Algebra (2021)

    duration: 3 hours; points: 50

    Problem 1 (points: 5)
    Suppose V is finite-dimensional and let `1 , `2 ∈ L (V). Prove that

    range(`1 ) ⊂ range(`2 ) ⇐⇒ ∃`0 ∈ L (V) such that `1 = `2 ◦ `0 .

    Problem 2 (points: 5+3+2=10)


    Let A ∈ Rn ×n be invertible. Apply Gram-Schmidt decomposition to the columns of A and show that we
    can decompose A as A = QR, where Q ∈ Rn ×n is an orthogonal matrix and R is an upper triangular matrix
    R whose diagonal elements are positive. By “upper triangular”, we mean that R is of the form
     
    × × × × ×
     × × × × 
    R= × × × .
     

    0 × ×
     
    ×

    Using the decomposition A = QR, explain


    (i) How you can solve Ax = b for some b ∈ Rm ?
    (ii) How you can verify whether det(A) belongs to the interval [−r, r ] for some given r > 0?

    Problem 3 (points: 5)
    Prove that P ∈ Rn×n is an orthogonal projector (onto a subspace of Rn ) if and only if P2 = P and P> = P.

    Problem 4 (points: 8+7=15)


    Let A, B ∈ Rn ×n be symmetric positive semidefinite matrices (all eigenvalues are nonnegative). Prove that

    A2 = B2 =⇒ A = B.

    (hint: use contradiction; assume that A 6= B and v be an eigenvector of A − B with nonzero eigenvalue).
    (ii) If C ∈ Rn ×n be defined as Ci j = Ai j Bi j , 1 ¶ i , j ¶ n , then show that C is symmetric positive semidefinite.

    Problem 5 (points: 5)
    Let A ∈ Rn ×n be symmetric and b ∈ Rn . Using just the eigendecomposition of A, prove that

    Ax = b has a solution ⇐⇒ ∀z ∈ Rn such that Az = 0 : z >b = 0.

    Problem 6 (points: 3+4+3=10)


    We have seen in a lecture that the PCA problem of finding an optimal k -dimensional subspace in Rn
    (k ¶ n) can be reduced to the following: Find orthonormal vectors u 1 , . . . , u k ∈ Rn such that the sum
    k
    X
    u >j Au j (1)
    j =1
    is maximized. Here, A ∈ Rn ×n is the symmetric positive semidefinite covariance matrix associated with
    the data. In particular, let the eigendecomposition of A be
    n
    X
    A= λi q i q >i , (2)
    i =1

    where λ1 ¾ λ2 ¾ · · · ¾ λn ¾ 0 and assume that that λ1 , . . . , λk are positive.

    (i) Substituting (2) in (1), show that we can write (1) as


    n
    X
    λi γi ,
    i =1
    Pn
    where γ1 , . . . , γn ∈ R are such that γi ∈ [0, 1] and i =1 γi = k.

    (ii) In particular, explain why the problem of maximizing (1) over all possible orthonormal vectors
    u 1 , . . . , u k ∈ Rn can be reduced to the following optimization problem:
    n
    X n
    X
    max λi γi subject to γi ∈ [0, 1], γi = k .
    γ1 ,...,γn ∈R
    i =1 i =1

    Prove that the maximum is λ1 +· · ·+λk which is attained when γ1 = · · · = γk = 1 and γk +1 = · · · = γn = 0.


    Pn Pn
    (hint: show that if γi ∈ [0, 1] and i =1 γi = k , then i =1 λi γi ¶ k ).

    (iii) Using (ii), show that (1) is maximized when u j = q j , 1 ¶ j ¶ k . That is, the optimal solution of the
    PCA problem is given by top eigenvectors of A.

    You might also like