Worksheet 23 (April 2)
DIS 119/120 GSI Xiaohan Yan
1 Review
METHODS AND IDEAS
Theorem 1. The row rank is equal to the column rank for any matrix. In fact
we have
dimRowpAq ` dimNulpAq “ n and dimColpAq ` dimNulpAq “ n,
where the former comes from the orthogonality, and the latter comes from the
Rank-Nullity theorem.
Remark 1.RowpAq “ NulpAqK,ColpAq “ NulpAT qK.
Remark 2. Another way to see the theorem is by row reduction. In fact, boththe row rank and the column rank are preserved by row reductions. (Why?) Sowe may reduce the theorem to the case of RREF. But in RREF both ranks areequal to the number of pivots.
Theorem 2. Otrhgonal matrices preserve the inner product. In other words,
given an orthogonal matrix U , we have
Ux ¨ Uy “ x ¨ y,@x,y P Rn.
Remark 3. In particular, this gives ||Ux|| “ ||x|| if we take x “ y.
2 Problems
Example 1. True or false.
( ) Let U be an orthogonal matrix, then detpUq “ 1.
( ) Let U be an orthogonal matrix and x a vector such that Ux and x arelinearly dependent, then Ux “ ˘x.
( ) If U is diagonal and orthogonal, then U must be an identity matrix.
1
dimRowla dimColas
pivot freevan
c Rhu c1pm
gconsiderAT RowCAT NullAT
ColinforAmxnanother
ideaofNff
normsangles
in ma47451455145
TutugnI'T.yi x.jo
detailLiu In
detail.atug detiuTug zsdetllD t1
Examine411 detiusa
F 41 detiut1
Eigenvaluesof I o thenux o soUI 5orthogonalmatrix I'ti u isamultipleofcanonlybe F isanevectorofus1 or 1 Counterexample 4 1 1 11411 111 UI's15
( ) The Gram-Schmidt process produces from a linearly independent set tu1, ¨ ¨ ¨ ,ukuan orthogonal set tw1, ¨ ¨ ¨ ,wku with the property that for each k,
spantw1, ¨ ¨ ¨ ,wiu “ spantu1, ¨ ¨ ¨ ,uiu,@i “ 1, 2, ¨ ¨ ¨ , k.
( ) For any two matrices A and B such that AB is well-defined,
rankAB § maxtrankA, rankBu.
Example 2. Find an example or disprove existence:a linear transformation T : R2 Ñ R2 that satisfies T 2 “ T but is not anorthogonal projection.
Example 3. Find the best fitting model in the form y “ ax2 ` bx ` c of thedata points
p´1, 1q, p0, 0q, p1, 1q, p2, 1q.
Example 4. Find the values of a, b, c, d, e, f, g such that U is an orthogonalmatrix: ¨
˝1 c e
a?22 f
b d g
˛
‚.
2
I
T mxn nxp mxp
tankABdimImHB
Be Itraink'Adementia
1mA1pm
mm widempotence aslantedprojection Takebasis pefbiyfj.baliHarbitrary oy yX
PITIFFE'sp p E Then T 1132 71132 1132 nw
direction Ji s b sprojx.AE B Tok X ly i s Xo TZTTo
Standardmatrixof 1 is lot TET TH TiTher TiImTcE
x iya E E p
Findtheleastsquaresolutionof RmdThismethodworkseventhoughthemodelis not alinear
model
I a b1C Itworksaslongasthemodel0 c dependslinearlyonthecoefficients
I a
ibtcl4atzbc.ATtxt ATb