Nonlinear Programming Exam Note 1
Nonlinear Programming Exam Note 1
If 1 and 2 are
in S, then convex combination 1 (12 must also belongs to S 0,1]
hyperplane S {: pt x } where p is a nonzero vector in Rn . If S, pt = , S {: pt (x- ) = 0}.
The convex hull of S, conv(S), is the collection of all convex combinations of S. xconv(S) iff x , =1, 0
conv(S) is the smallest convex set containing S. conv(S) is the intersection of all convex sets containing S.
The convex hull of a finite number of points 1 ,...,+1 in Rn is called a polytope.
If 1 ,...,+1 are affinely independent, which means that 2 1 ,, +1 1 , are linearly independent.
Note: 1 ,..., are linearly independent if 1 1 + + we have =0for all i.
If 1 ,...,+1 are affinely independent, then conv 1 ,...,+1 is called a simplex having vertices 1 ,...,+1 .
S is a set in Rn ,if Sconv(S),then Sconv(1 ,...,+1 ), and x can be written as x , =1, 0
A point x is in the closure of S, denoted by cl S, if S () >0.For any neighborhood of x, there is some
element of S in the neighborhood. cl S=int S S is the smallest closed set containing S. S=[0,1){2}cl
S=[0,1]{2}
A point x is in the interior of S, denoted int S, if ()S for some > 0.
If S=cl S, S is called closed. If S=int S, S is called open. S=[0,1]{2}= cl Sclosed.
A set S is closed iff it contains all its boundary points. int S S S , while SS int S necessarily.
A set is open iff it does not contain any of its boundary points. S =
A set S is closed iff for any convergent sequence of points { } contained in S with limit point , we also have xS.
A point x is in the boundary of S, S , if N (x) contains at least one point in S and one point not in S for >0.
S is a nonempty, compact set, and f : SR is continuous on S.Then the problem min{f x: xS}attains its minimum.
S is a nonempty closed convex set in Rn and yS.(1)There exists a unique point S with the minimum distance from
y iff (y )t (x ) 0 S (2)There exists a nonzero vector p and a scalar such that pt y and pt x for
xS . ( hyperplane H {: pt x }separate S1 and S2 if pt xx1 and pt x x2 )
S1 Ax0, t >0, x Ax<0, x Ax0, t >0, x0, Ax0,B =0, t >0, x l n
x
S2 t = , y0, y t =0, y0, y t c, y0, y t + t = ,y0, y,z ,
H {x : pt (x ) 0} is a supporting hyperplane of S at if either pt (x )0 or pt (x )0 for each xS
Let S be a nonempty convex set in Rn , (1)and let S. Then there exists a hyperplane that supports S at ; that is,
there exists a nonzero vector p such that pt (x ) 0 for each xcl S.
(2)and xint S . Then there is a nonzero vector p such that pt (x ) 0 for each xcl S.
(3)and let yint conv(S) . Then there exists a hyperplane that separates S and y.
(4)and let Sconv(S) . Then there exists a hyperplane that supports S at .
Let S1 and S2 be nonempty convex sets in Rn and suppose that either (1)S1S2 is empty. or(2) int S is not empty and
that S1 intS2 is empty Then there exists a hyperplane that separates S1 and S2 ; that is, there exists a nonzero vector p
in Rn such that inf{pt x: xS1}sup{pt x: xS2}, which means pt 1 pt 2 for all 1 S1 and 2 S2.
Let S1 and S2 be closed convex sets, and suppose that S1 is bounded. If S1S2 is empty, there exists a hyperplane that
strong separates S1 and S2; that is, there exists a nonzero p and 0 such that inf{pt x: xS1} +sup{pt x: xS2}
Cone:A nonempty set C in Rn with vertex zero if xC implies that xC 0. If C is convex, C is a convex cone.
The polar cone of S, S*, is given byp:pt x 0 xS. If S is empty, S* is Rn . A closed convex cone x : Ax 0
A polyhedral set is a closed convex set, the intersection of a finite number of closed half-spaces;S={x: }
A vector xS is an extreme point of a nonempty convex set S in Rn if x1 (12 (0,1) implies x1 2
A nonzero vector d is a direction if xS, x+dS 0.Two directions d1 and d2 are distinct if d1 d2 >0.
(1)If the set is bounded, then it has no directions. (2)The number of extreme points and extreme directions of S is finite.
A direction d in S is an extreme direction if d 1 d1 2d2 for 1,2 0, then d1=d2 for some > 0.
Feasible Direction: S {x : Ax b, x0}, A(m n)is of rank m. (1)d0 is a direction of S iff Ad=0 for d0.
(2)xS iff x + j and j=1, , j0S has at least one extreme direction/point if it is
unbounded/nonempty.
(4)To have a finite optimal solution iff dj0 for all extreme directions of the feasible region
Minimize t subject to Ax b, x0, = , and , 0. B N b, = B1 B1 N
t = + = (B 1 B 1 N ) + = B 1 +( B 1 ) . If ( B1 ) 0, = B1
D: Maximize t y subject to t y c, y unrestricted.
Weak duality result: t ty for any feasible solution x to P and any feasible solution y to D.
Strong duality result: If both P and D are feasible, they both have optimal solutions with the same objective value.
Unbounded-infeasible relationship: If P is unbounded, D is infeasible. If D is unbounded, P is infeasible.
If D is infeasible, P is unbounded or infeasible. If P is infeasible, D is unbounded or infeasible.
Let be a primal feasible solution, and be a dual feasible solution. and are optimal to P and D iff =0
j=1,..n where =( , =ct
1 , ) is the vector of slack variables in the dual constraints for the dual solution .
If > 0,then =0. The equality of the th constraint in dual holds. If > 0,then =0, which implies =0.
A set F R is closed if the complement of F, R \ F, is open. [0, 1] {(-,0) and (1,) open}, [0, ) {(-
,0)open}
(0, 1] is neither open nor closed. It is not open, because x = 1 is contained in the set, but every neighborhood of
x = 1 is not contained in the set. It is not closed, because its complement - ,0] and (1, ) is not open at x = 0.
A compact set must be both closed and bounded. [0,1] is compact.
X is a limit of ( ) iff for each > 0 there exists K N s. t. n > k() | | <
For each > 0, choose K N such that K > 1/, when n > k, | 0| = |1/n 0| 1/k <
A set S is closed iff convergent sequence of points ( ) contained in S with limit point X XS
S=[0,1] is closed, ( )=(1/2)^n 0, 0S S=(0,1] is not closed because 0 is not in S
S is not closed, there exists ( ) n s.t. S but x does not belong to S.
A set X is convex if for any x and x ' belonging to X and any [0,1] , x+(1-)x 'X .
A correspondence C is called (usc), if the graph of the correspondence {(x, y) : yC(x)} is closed.
A fixed point of r is a s.t. r(),i.e., must be one of player is best responses to
Kakutanis fixed point theorem Let S n be a compact and convex set. Let C be a correspondence from S
into
itself that is usc and convex valued. Then, there is an S such that C( ) .C: best response. S:solution space
Sufficient condition for r: have a fixed point: 1. is compact,convex and nonempty 2.r() is not empty .
n ) (, ) with n r(n ), then r()
3. r() is convex . 4.r(.) has a closed graph: If sequence (n ,