《Matrix Methods in Data Analysis, Signal Processing, and Machine Learning》MIT Course Number:18.065课程习题答案
LEC3. Orthonormal Columns In Q Give Q’Q= I
Problems of Lecture 3 (from textbook Section I.5)
2 Draw unit vectors u and v that are not orthogonal. Show that ω = v − u ( u T v ) \omega = \textbf{v} − \textbf{u} (\textbf{u}^T\textbf{v}) ω=v−u(uTv) is orthogonal to u (and add w to your picture).
4 Key property of every orthogonal matrix : ∣ ∣ Q x ∣ ∣ 2 = ∣ ∣ x ∣ ∣ 2 ||Qx||^2 =||x||^2 ∣∣Qx∣∣2=∣∣x∣∣2 for every vector x.More than this, show that ( Q x ) T ( Q y ) = x T y (Qx)^T (Qy) = x^Ty (Qx)T(Qy)=xTy for every vector x and y. So lengths and angles are not changed by Q. Computations with Q never overflow !
6 A permutation matrix has the same columns as the identity matrix (in some order). Explain why this permutation matrix and every permutation matrix is orthogonal :
P
=
[
0
1
0
0
0
0
1
0
0
0
0
1
1
0
0
0
]
h
a
s
o
r
t
h
o
n
o
r
m
a
l
c
o
l
u
m
n
s
s
o
P
T
P
=
∗
∗
∗
‾
a
n
d
P
−
1
=
∗
∗
∗
‾
.
P= \begin{bmatrix}0&1&0&0\\0&0&1&0\\0&0&0&1\\1&0&0&0\\ \end{bmatrix} has \ orthonormal \ columns \ so \ P^TP=\underline{ ***} \ and \ P^{-1}=\underline{***}.
P=⎣⎢⎢⎡0001100001000010⎦⎥⎥⎤has orthonormal columns so PTP=∗∗∗ and P−1=∗∗∗.
When a matrix is symmetric or orthogonal, it will have orthogonal eigenvectors.
This is the most important source of orthogonal vectors in applied mathematics.
Solutions to Lecture 3
2 v is separated into a piece u ( u T v ) u(u^Tv) u(uTv) in the direction of u and the remaining piece ω = v − u ( u T v ) \omega = \textbf{v} − \textbf{u}(\textbf{u}^T\textbf{v}) ω=v−u(uTv) perpendicular to u. Check u T w = u T v − ( u T u ) ( u T v ) = 0 u^Tw = u^T v − (u^Tu)(u^T v) = 0 uTw=uTv−(uTu)(uTv)=0
4 Check ( Q x ) T ( Q y ) = x T Q T Q y = x T y (Qx)^T (Qy) = x^TQ^TQy = x^Ty (Qx)T(Qy)=xTQTQy=xTy : Angles are preserved when all vectors are multiplied by Q. Remember x T y = ∣ ∣ x ∣ ∣ ∣ ∣ y ∣ ∣ c o s θ = ( Q x ) T ( Q y ) x^Ty = ||x|| \ ||y||cosθ = (Qx)^T (Qy) xTy=∣∣x∣∣ ∣∣y∣∣cosθ=(Qx)T(Qy): same θ!
6 Every permutation matrix has unit vectors in its columns (single 1 and (n − 1) zeros). Those columns are orthogonal because their 1’s are in different positions. P T P = I ‾ a n d P − 1 = P T ‾ P^TP=\underline{I} \ and \ P^{-1}=\underline{P^T} PTP=I and P−1=PT