Page 138 -
P. 138

unimodal  function  A  function  which  has
                                                           one mode (usually a maximum, but could mean a
                                                           minimum, depending on context). If f is defined
                                 U                         on the interval [a, b], let x be its mode.  Then,
                                                                                 ∗
                                                                                       ∗
                                                           f  strictly  increases  from  a  to  x and  strictly
                                                                         ∗
                                                           decreases from x to b (reverse the monotonicity
                                                                          ∗
                                                           on each side of x if the mode is a minimum).
                  unbounded  mathematical  program   The   (For  line  search  methods,  like  Fibonacci,  the
                  objective is not bounded on the feasible region  mode could occur in an interval, [a , b ], where
                                                                                          ∗
                                                                                       ∗
                  (from above, if maximizing; from below, if min-  f  strictly increases from a to a , is constant (at
                                                                                     ∗
                  imizing).  Equivalently, there exists a sequence  its global max value) on [a , b ],  then strictly
                                                                                  ∗
                                                                                     ∗
                                      k              k
                  of feasible points,  say {x } for which {f (x )}  decreases on [b , b].)
                                                                        ∗
                  diverges to infinity (minus infinity, if minimiz-
                  ing).
                                                           unimodular  matrix   A  nonsingular  matrix
                                                           whose determinant has magnitude 1.  A square
                  unconstrained  mathematical  program  One
                                                           matrix is totally unimodular if every nonsingular
                  with no constraints (can still have X be a proper  submatrix from it is unimodular.  This arises in
                           n
                  subset of R , such as requiring x to be integer-
                                                           (linear) integer programming because it implies
                  valued.)
                                                           a basic solution to the LP relaxation is integer-
                                                           valued (given integer-valued right-hand sides),
                  unconstrained  optimization  Taken  liter-
                                                           thus obtaining a solution simply by a simplex
                  ally,  this  is  an  unconstrained  mathematical
                                                           method.  An example of a totally unimodu-
                  program.  However, this phrase is also used in
                                                           lar matrix is the node-arc incidence matrix of
                  a  context  that  X  could  contain  the  strict  inte-
                                                           a network, so basic solutions of network flows
                  rior,  with  constraints  of  the  form  g(x) <  0,
                                                           are integer-valued (given integer-valued supplies
                  but the mathematical program behaves as uncon-
                                                           and demands).
                  strained. This arises in the context of some algo-
                  rithm design, as the solution is known to lie in the
                  interior of X, such as with the barrier function.  unimolecular  See molecularity.
                  uncountably  infinite  set  An  infinite  set  unisolvence  A set of functionals on a finite
                  which is not denumerably so. See also cardinal-  dimensional vector space V is called unisolvent,
                                                                                 h
                  ity, countable set, denumerably infinite set, finite  if it provides a basis of the dual space of V .
                                                                                                h
                  set, and infinite set.                    Unisolvence is an essential property of degrees
                                                           of freedom in the finite element method.
                  undirected edge  See edge.
                                                           unit    An identity element.
                  unified  atomic  mass  unit  Non-SI  unit  of
                  mass (equal to the atomic mass constant), defined
                  as one twelfth of the mass of a carbon-12 atom  unit circle  A circle of radius 1. Usually, the
                  in its ground state and used to express masses  term refers to the circle of radius 1 and center 0
                                                           in the complex plane ({z : |z|= 1}).
                  of  atomic  particles,  u =  1.660  5402(10) ×
                  10 −27  kg.
                                                           unitary group    An automorphism α    :
                  uniformly bounded   Referring to a family F  C → C is called unitary if, once a basis E has
                                                             m
                                                                   m
                                                                                              i
                  of functions such that the same bound holds for  been chosen in C , the matrix U representing
                                                                                      j
                                                                         m
                                                                                      i
                  all functions in F. For example,                                     j
                                                           the automorphism by α(E ) = U E satisfies
                                                                                 j
                                                                                          j
                                                                                       i
                                                                    †
                                 f (x) ≤ µ                 U  −1  = U , i.e., if the inverse of U coincides
                                                           with the transpose of the complex conjugated
                  with the same µ, for all f ∈ F.          matrix.
                  c
           © 2003 by CRC Press LLC
           © 2003 by CRC Press LLC
   133   134   135   136   137   138   139   140   141   142   143