Back to Singular Value Decompositions

Activity 7.4.4

Avg difficulty 2.5

Let's suppose that a matrix AA has a singular value decomposition A=UΣVTA=U\Sigma V^T where U=[u1u2u3u4],Σ=[2000050000000],V=[v1v2v3].U=\begin{bmatrix} \uvec_1 & \uvec_2 & \uvec_3 & \uvec_4 \end{bmatrix},\hspace{10pt} \Sigma = \begin{bmatrix} 20 & 0 & 0 \\ 0 & 5 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{bmatrix},\hspace{10pt} V=\begin{bmatrix} \vvec_1 & \vvec_2 & \vvec_3 \end{bmatrix}.

  1. What is the shape of AA; that is, how many rows and columns does AA have?

  2. Suppose we write a three-dimensional vector x\xvec as a linear combination of right singular vectors: x=c1v1+c2v2+c3v3.\xvec = c_1\vvec_1 + c_2\vvec_2 + c_3\vvec_3\text{.} We would like to find an expression for AxA\xvec.

    To begin, VTx=[v1⋅xv2⋅xv3⋅x]=[c1c2c3]V^T\xvec = \threevec{\vvec_1\cdot\xvec} {\vvec_2\cdot\xvec} {\vvec_3\cdot\xvec} = \threevec{c_1}{c_2}{c_3}.

    Now ΣVTx=[2000050000000][c1c2c3]=[20c15c200]\Sigma V^T \xvec = \begin{bmatrix} 20 & 0 & 0 \\ 0 & 5 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{bmatrix}\threevec{c_1}{c_2}{c_3} = \cfourvec{20c_1}{5c_2}00.

    And finally, Ax=UΣVTx=[u1u2u3u4][20c15c200]=20c1u1+5c2u2A\xvec = U\Sigma V^T\xvec = \begin{bmatrix} \uvec_1 & \uvec_2 & \uvec_3 & \uvec_4 \end{bmatrix} \cfourvec{20c_1}{5c_2}00 = 20c_1\uvec_1 + 5c_2\uvec_2.

    To summarize, we have Ax=20c1u1+5c2u2A\xvec = 20c_1\uvec_1 + 5c_2\uvec_2.

    What condition on c1c_1, c2c_2, and c3c_3 must be satisfied if x\xvec is a solution to the equation Ax=40u1+20u2A\xvec=40\uvec_1 + 20\uvec_2? Is there a unique solution or infinitely many?

  3. Remembering that u1\uvec_1 and u2\uvec_2 are linearly independent, what condition on c1c_1, c2c_2, and c3c_3 must be satisfied if Ax=0A\xvec = \zerovec?

  4. How do the right singular vectors vi\vvec_i provide a basis for Nul(A)\nul(A), the subspace of solutions to the equation Ax=0A\xvec = \zerovec?

  5. Remember that b\bvec is in Col(A)\col(A) if the equation Ax=bA\xvec = \bvec is consistent, which means that Ax=20c1u1+5c2u2=bA\xvec = 20c_1\uvec_1 + 5c_2\uvec_2 = \bvec for some coefficients c1c_1 and c2c_2. How do the left singular vectors ui\uvec_i provide an orthonormal basis for Col(A)\col(A)?

  6. Remember that rank⁡(A)\rank(A) is the dimension of the column space. What is rank⁡(A)\rank(A) and how do the number of nonzero singular values determine rank⁡(A)\rank(A)?

Austin, Understanding Linear Algebra, §7.4 Singular Value Decompositions, Activity 7.4.4

AI Tutor

Adapted from Understanding Linear Algebra © 2023 David Austin. Licensed under CC BY 4.0. Source: Understanding Linear Algebra. Imported at revision 3e6dced.