OPTIMIZATION THEORY
++SIMPLEX METHOD



    |
    |
    |\
    | \
    |  \
    |   \
    |-___\__
    |     \ ---___________
    |      \ 
    +------------------------
        /\
        Polytope


B+W < 350
2B+W< 600
B+3W< 900

(1947 George Dantzig)
LP (Lin Prog) Solution always @ corner points on Boundary of Polytope
Algebraic Topology - polytope (Feasible region) = "SimpleX"

Simplex Method 
    1. Find a vertex in polytope & find value  @ this pt
    2. Examine each edge attached to this vertex
    3. Move alone edge w/ best improvement of criterion
    4. Repeat 2-3




Graph Theory (1736 Leonhard Euler)
Nodes 
    a. Beginning - odd incident degree
    b. End - odd incident degree
    c. Intermediate Node - even incident degree
     .--X--.
    / _/    \
   |_/       |
   X_________X
   |\        |
   \ \_      /
    \_ \   _/
      --X-- 



Min Cut-Max Flow Theorem
   V* = Max steady state flow of s& t (source- sink)
   c(P*,~P*) = capacity of minimum cut
    -> V* = c(P*,~P*)
cut = node sets P & ~P,  P w/ source node, ~P w/ sink node
Capacity = c(P,~P) = sum of edges w/ one node in P and the other endpt in ~P

             __   
            /  \  
          ->|  |- 
         /  \__/ \ 
        /    |\   \
       /     ||    \
      /      \|     \
  __ /       __      \  __     
 /  \       /  \       /  \ 
 |  | ----> |  |       |  |
 \__/       \__/       \__/ 
     \       | 
      \      | 
       \     \ 
        \    __           
         \  /  \    
          ->|  |    
            \__/    

Gradient Method/Routing Networks
1. 1931 W. Gropius (SAR/OSR) Housing Block
P=alx/b 
A=l(a+b)
I=3x/s
A/P = b(aI+3x) / alx
OSR = bsl/alx = bs/ax = sl/P
SAR -b/x = A/P -  b/x = sl/P = OSR

2. HILL CLIMB 
  Move in direction of steepest ascent/Gradient
Abraham wald- dynamic programming
3. Network routing (c.1950 R. Bellman)
Optimal value func - minimal cost tour city k to city N
optimal policy func j*(k) next city if at city k
4. Money allocation 
principle of optimality

I3(X) = Max return of 3 investments w/ capital $X
I3(X)=max(0.xc.X)[g3(xc)+I2(X-xc)]
g1(y)=return of investment 1 w/ $y



S Home