Skip to main content

Proof BlocksLinear Algebra II

All proofs
ProofAnvil Practice · Linear Algebra II

Proof Blocks

The steps of a proof are mixed up, with a few extra steps that don't belong. Put the right steps in order.

A college study tested this kind of practice. After reading the same notes, students learned about as much as those who wrote proofs from scratch, in about a quarter of the time. Read the study

LetA=(210120)∈M2×3(R).A=\begin{pmatrix}2&1&0\\1&2&0\end{pmatrix}\in M_{2\times3}(\mathbb R).1. Compute AATAA^T and ATAA^TA. Find the eigenvalues of both, and prove that the nonzero eigenvalues coincide. Extract the singular values σ1≥σ2≥σ3≥0\sigma_1\ge\sigma_2\ge\sigma_3\ge0 of AA, including the structural zero forced by the shape.

2. Construct orthonormal eigenbases of AATAA^T and of ATAA^TA that are compatible in the sense Avi=σiuiAv_i=\sigma_i u_i whenever σi>0\sigma_i>0. Write a full SVD A=UΣVTA=U\Sigma V^T with U∈O(2)U\in O(2), V∈O(3)V\in O(3), and Σ\Sigma of size 2×32\times3 having the singular values on its main diagonal.

3. Prove that the ordered list of singular values is uniquely determined by AA (even though UU and VV are not). Compute rank⁡A\operatorname{rank}A, ∥A∥2\|A\|_2, and ∥A∥F\|A\|_F.

4. Let A1=σ1u1v1TA_1=\sigma_1 u_1 v_1^T. Write A1A_1 as an explicit 2×32\times3 matrix, prove that rank⁡A1=1\operatorname{rank}A_1=1, and prove that∥A−A1∥2=σ2.\|A-A_1\|_2=\sigma_2.Explain why no rank-at-most-one matrix can have a strictly smaller operator-norm error.

Do not treat the already nearly diagonal appearance of AA as an SVD, and do not omit the third right singular vector.

Opening the proof…

Choose an activity

Join to unlock the full course and save your place.

Honors has roughly twice as many lines to arrange.
  1. Path position 014

    Compute a full singular-value decomposition of a rank-two 2×32\times3 matrix, prove that the singular values are uniquely determined, and identify the Eckart–Young rank-one approximant together with its exact operator-norm error.

  2. Path position 006

    Recover nilpotent Jordan multiplicities from successive kernel increments and distinguish them from ranks.

  3. Path position 009

    Show that characteristic polynomial, minimal polynomial, and geometric multiplicity can leave two Jordan profiles unresolved, then use kernels of powers as similarity invariants to distinguish and reconstruct them.

  4. Path position 010

    Construct the right and left polar decompositions of an explicit invertible real matrix from the positive-definite square root, prove uniqueness of the factors, and keep the same unitary factor on both sides.

  5. Path position 012

    Prove the geometric characterization of linear least squares, compute the orthogonal projection of a concrete vector onto a plane spanned by orthogonal generators, and verify Pythagoras together with uniqueness from full column rank.

5 sample Linear Algebra II Proof Blocks activities are ready.