Projections & Decompositions - Linear Algebra

  1. What is projWy\text{proj}_W\mathbf{y} when y\mathbf{y} lies in W⊥W^\perp?
    0\mathbf{0}. All of y\mathbf{y} is in the orthogonal part of the decomposition.
  2. If y=y^+z\mathbf{y} = \hat{\mathbf{y}} + \mathbf{z} with y^\hat{\mathbf{y}} in WW and z\mathbf{z} in W⊥W^\perp, how are the squared norms related?
    ∥y∥2=∥y^∥2+∥z∥2\|\mathbf{y}\|^2 = \|\hat{\mathbf{y}}\|^2 + \|\mathbf{z}\|^2, the Pythagorean theorem, since the two parts are orthogonal.
  3. If PP is the matrix of orthogonal projection onto WW, what does I−PI - P do?
    It projects orthogonally onto W⊥W^\perp, returning the component y−projWy\mathbf{y} - \text{proj}_W\mathbf{y}.
  4. For WW the xyxy-plane in R3\mathbb{R}^3, what is the orthogonal decomposition of y=(2,−3,7)\mathbf{y} = (2, -3, 7)?
    (2,−3,0)+(0,0,7)(2, -3, 0) + (0, 0, 7). The first part lies in WW and the second lies in W⊥W^\perp, the zz-axis.
  5. Why does applying an orthogonal projection PP twice give the same result as applying it once?
    After one projection the vector already lies in WW, and projection leaves vectors in WW unchanged. So P2=PP^2 = P.
  6. What is the trace of the orthogonal projection matrix onto a kk-dimensional subspace of Rn\mathbb{R}^n?
    kk. The trace is the sum of the eigenvalues: kk ones and n−kn - k zeros.
  7. If ∥y∥=13\|\mathbf{y}\| = 13 and ∥projWy∥=5\|\text{proj}_W\mathbf{y}\| = 5, what is ∥y−projWy∥\|\mathbf{y} - \text{proj}_W\mathbf{y}\|?
    1212. The pieces are orthogonal, so 132=52+12213^2 = 5^2 + 12^2.
  8. By the Orthogonal Decomposition Theorem, how can each y\mathbf{y} in Rn\mathbb{R}^n be written relative to a subspace WW?
    Uniquely as y=y^+z\mathbf{y} = \hat{\mathbf{y}} + \mathbf{z} with y^\hat{\mathbf{y}} in WW and z\mathbf{z} in W⊥W^\perp. Here y^=projWy\hat{\mathbf{y}} = \text{proj}_W\mathbf{y}.
  9. What is projWy\text{proj}_W\mathbf{y} when y\mathbf{y} already lies in WW?
    y\mathbf{y} itself. Its component in W⊥W^\perp is 0\mathbf{0}.
  10. What are the only possible eigenvalues of an orthogonal projection matrix?
    00 and 11. Vectors in WW have eigenvalue 11 and vectors in W⊥W^\perp have eigenvalue 00.
  11. Besides P2=PP^2 = P, what property of its transpose does every orthogonal projection matrix PP have?
    It is symmetric: PT=PP^T = P. A matrix with both properties is an orthogonal projection.
  12. What is the rank of the orthogonal projection matrix onto a kk-dimensional subspace WW?
    kk. Its column space is exactly WW.

Study these flashcards