s spano · 2020. 11. 26. · spanos, the above construction of an orthogonal basis is called...

15
Lecture 19: Recall : n < T.fi ) I ; - - T e Spank T , .ir , - sin } ) T - - I a Tritt - i = I orthogonal S ' G - S process . , S " { Ii , Is . . . inn } ms { t.TT } " orthogonal Basis , L . I . { II ! Yin - ni us Spano 's - . Spanos ,

Upload: others

Post on 29-Jan-2021

1 views

Category:

Documents


0 download

TRANSCRIPT

  • Lecture 19:

    Recall : n < T.fi ) I ;- -T e Spank T , .ir , . -

    . sin } ) ⇒ T - - I a Tritt① - i = I

    orthogonal

    S'

    ② G - S process . , S"

    { Ii , Is . . . inn }ms { t.TT }

    " orthogonalBasis , L .

    I.

    { II!Yin - ni. usSpano

    's -. Spanos ,

  • The above construction of an orthogonal basis is called

    Gram - Schmidt process .

  • Example :

    Consider V = PUR ) equipped with the inner product

    < f, g 7 =L

    '

    ,

    ft gets dt .Let p= { I ,

    x,

    x ? - - .,

    X"

    , -

    - - } be standard ordered basis

    for PHR ) .

    Take T , = I . O ," "

    Then : Tz = x - CK ,T

    . > -- V , = xHv.

    no2

    is . . i . a ;÷9%, -ax3v in = a- Iz

    11 Til 3

  • i. as .a÷i . . '= X

    '- ZX and So on i - - -

    for PUR)

    This produces an orthogonal basis{ Ti

    ,Tz

    ,.

    .

    },

    whose

    elements are called Legendre polynomial .

  • Corollary : Let V be a non - zero finite - diminner product space .

    Then, V has an orthonormal

    basis p = { I , , . . ,Tn3Sit .

    HI EV,

    wehave = I =

    ,

    Ti > Tri

    G - olla - y : Let V be a non- zero finite - dim inner product space

    with an orthonormal basis p - {I

    , ,Tz , . . , In } . Let T be

    a linear operator on V . Let A = ET ]p . Then : Aij-LTCTjs.vn .

    Proof : [ TIP =/ . . . fityjgyp.

    ..

    )TED -- it! iTi

    c- Aij

  • Orthogonal complement

    Def : Let S be a non - empty subset of an inner product

    Space V.

    The orthogonal complement of S is definedas :

    St Eet { Ie V . . s I , I >= o for V-J e S }

    -

  • Proposition : Let V be an inner product space and Wc ✓ a

    finite - dim subspace of V . Then i tty EV , I ! it E W and I C- Wt

    such that J -- It E .

    Furthermore , if {I

    , .ir , . . . step} is an orthonormalbasis for W

    ,

    hthen : I = -2 Bi

    i =L

    The vector view is called the orthogonal projectionof

    j on W .

    V

  • Proof : Given I e V , we set indef

    ¥45 , Ti > Ji EWand I ihty - I . Then : I - - It I .

    a

    W Tnt ? ?Now , LI

    ,Jj > = CJ - I , Tj > = a J , Tj > - < I,Tj >

    = e5,

    Tj > -

    [email protected].

    o

    . .

    " in

    .

    '

    . I e Wt

  • For uniqueness , suppose IT'

    EW and I'

    E Wt Such that :

    I - - It I = I 't E' ⇒ I - I

    '= E ' - I E Wnwt

    in

    a a A wta at w ht ww w

    Claim : w n wt = 3 T }

    P Take J e wnwt . Then : s I , To >= 0

    a a

    W wt⇒ in -6

    This implies : it -I ' = I ' - I -_ T

    ⇒ I = I'

    and I -_E'

    .

  • Corollary : With notations asabove , then :

    115 - Ill 7115 - Ill forVIEW

    A

    and equality hFIdsFffI -_ ITW

    yRe-make. Orthogonal projection is the gun . gonalvector in W closest to I . proj

    .

  • : Let I e W . Then : J - - uit I ⇒ E-

    - I -8a

    w"

    wt

    ily-5112=11uitz - IIT =L xtI,@tE >" war It "w "w Ta

    = L T - I,

    I - I > ten - I. I > + CE , I - I > t SEE >11 It

    . il

    o O

    = Itu - Ill'

    t HER z HEH'

    = 115 - FIT

    The equality holds iff Hui - Ill'

    -_ o iff I -- I .

  • Proposition : Suppose S = { I , ,Tz , . - . , Th } is anorthonormal

    pet in an n - dimensionalinner product space V . Then :

    ( a ) S can extended to anorthonormal basis

    { Ji , -Vz , . - , Th , Treat , , . . . ,In } for V .

    (b) If W= span I S ) , then S ,= { They , . . ,Tn3 is

    an orthonormal basis for Wt

    (c) If W is any subspaceof V

    ,then :

    dim I V ) = dimLw ) t dim C

    Wt )

  • Proof : ( a ) we first extends to abasis i

    a{ Ii

    , . . ,

    Th,

    The,

    -. .

    ,In } for V .

    -

    L . I .

    Then, we apply the G - S process

    to this basis .

    '

    .

    '

    S is orthonormal , .'

    ,Ti

    ,. .

    ,Tres remains the same

    during the G - S process .

    So,

    this process gives an orthonormalbasis for V of

    the form { Ti , Ja , . . , Th , The , , . . ,In }

    --

    unchangednew

  • (b) Note :. S , = { The , . . ,Tn3 is linearly independent (

    '

    i

    '

    it is

    part of• S

    ,c Wt ( We span i , . . ,Th) )

    the basis )

    ⇒ Spanish cwt

    Suffices to show = Span C S , )= Wt

    .

    For Ie V,

    we have I = ¥45IiIf I Ewf then C I , Ii > = o for 5=1,2 ,

    - - ik

    n

    Then : I - I a I , Tri > I ; E Span LSDj= htt

    ! . Wtc Span I s , ) ⇒ w 'T Span CS , )

  • Kc ) For any W , choose anorthonormal basis { Ji

    , . . ,Trees

    for W and extendit to an orthonormal basis

    { Tl , - - , Th , Berti , - . ,En } for V .

    Then :

    dim LV ) = h= kt C n - h )

    = dinlw ) t dim I Wt )