Page 90 - Advanced Linear Algebra
P. 90

74    Advanced Linear Algebra



             )
            1 If =~ : l ;   then
                                                :Á; b    ; Á: ~
             )

            2 If  ~     :Á;  then

                                 im²³ ~ :   and  ker ²³ ~ ;

               and so
                                      =~ im ² ³ l ker ² ³


               In other words,   is projection onto its image along its kernel. Moreover,

                                    # im ² ³  ¯      # ~#

             )

                     B
            3   If  ²= ³  has the property that
                              =~ im ² ³ l ker ² ³      O   and  ~   im²³





               then   is projection onto im²³  along ker ²³ .…
            Projection operators are easy to characterize.
                                       B

            Definition A linear operator  ²= ³  is idempotent  if        .   ~  …
            Theorem 2.22 A linear operator    B ²= ³  is a projection if and only if it is
            idempotent.
            Proof. If  ~        , then for any   :Á;     :   and    !  , ;
                                            ²  b !³~  ~ ~ ²  b !³

            and so          ~  . Conversely, suppose that   is idempotent. If #     im ²      ³  ker ²    q  , ³
            then #~ %  and so



                                   ~ #~       %~ %~#
            Hence im² ³ q ker ² ³ ~ ¸ ¹ . Also, if #  =  , then


                             #~²# c #³ b # ker  ² ³ l im ² ³




            and  so  =~ ker ² ³ l im ² ³ . Finally,  ² %³ ~         % ~ %   and so  O im ²³  ~  .






            Hence,   is projection onto im²³  along ker ²³ .…



            Projections and Invariance
                                                                    B

            Projections can be used to characterize invariant subspaces. Let  ²= ³  and
                                                                  :
              :
                                                             ;
            let   be a subspace of  . Let  ~  =        :Á;   for any complement   of  . The key is
            that the elements of   can be characterized as those vectors fixed by  , that is,
                             :
   85   86   87   88   89   90   91   92   93   94   95