Page 392 - 35Linear Algebra
P. 392

392                                                                                Movie Scripts








                                      3                                 2
                                     Z 2                           or Z 2






                            Now lets consider a linear transformation
                                                                 3     2
                                                            L : Z −→ Z .
                                                                 2     2
                            This must be represented by a matrix, and lets take the example

                                                                     
                                                      x               x
                                                             0  1  1
                                                   L   y   =         y   := AX .
                                                             1  1  0
                                                       z                z
                            Since we have bits, we can work out what L does to every vector, this is listed
                            below
                                                                   L
                                                          (0, 0, 0) 7→ (0, 0)
                                                                   L
                                                          (0, 0, 1) 7→ (1, 0)
                                                                   L
                                                          (1, 1, 0) 7→ (1, 0)
                                                                   L
                                                          (1, 0, 0) 7→ (0, 1)
                                                                   L
                                                          (0, 1, 1) 7→ (0, 1)
                                                                   L
                                                          (0, 1, 0) 7→ (1, 1)
                                                                   L
                                                          (1, 0, 1) 7→ (1, 1)
                                                                   L
                                                          (1, 1, 1) 7→  (1, 1)
                            Now lets think about left and right inverses. A left inverse B to the matrix
                            A would obey
                                                               BA = I

                            and since the identity matrix is square, B must be 2 × 3. It would have to
                                                                        3
                            undo the action of A and return vectors in Z to where they started from. But
                                                                        2
                                                                     3                                   2
                            above, we see that different vectors in Z are mapped to the same vector in Z
                                                                     2                                   2
                            by the linear transformation L with matrix A. So B cannot exist. However a
                            right inverse C obeying
                                                               AC = I
                                                                                  2                 3
                            can. It would be 2×2. Its job is to take a vector in Z back to one in Z in a
                                                                                  2
                                                                                                    2
                            way that gets undone by the action of A. This can be done, but not uniquely.
                                                      392
   387   388   389   390   391   392   393   394   395   396   397