Scribd
Upload
2 views

SimplexNotes

The Simplex Method is an iterative approach for solving linear programming problems, utilizing both geometric and algebraic perspectives. It involves starting from a corner point of a polytope and moving along edges to find optimal solutions, while also introducing slack variables to reformulate the problem. The method identifies feasible solutions through the intersection of hyperplanes defined by constraints, ultimately leading to basic feasible solutions that correspond to corner points of the feasible region.

Uploaded by

ezekielakpejeriosare
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
2 views

SimplexNotes

The Simplex Method is an iterative approach for solving linear programming problems, utilizing both geometric and algebraic perspectives. It involves starting from a corner point of a polytope and moving along edges to find optimal solutions, while also introducing slack variables to reformulate the problem. The method identifies feasible solutions through the intersection of hyperplanes defined by constraints, ultimately leading to basic feasible solutions that correspond to corner points of the feasible region.

Uploaded by

ezekielakpejeriosare
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

The SIMPLEX METHOD is an iterative method for solving linear programming problems.

We discuss the method from both geometric and algebraic vantage points.

The Simplex Method reduces things to solving equivalent systems of linear equations.

We start with an LP in standard form


maximize z = cT x
subject to Ax ≤ b, x ≥ 0.

The bounds (faces) of the polytope come from the n equalities xi = 0, i = 1, · · · , n (the
original decision variables), and also from the m equalities found in the matrix equation
Ax = b:
x1 = 0
x2 = 0
..
.
xn = 0
a11 x1 + a12 x2 + · · · + a1n xn = b1
a21 x1 + a22 x2 + · · · + a2n xn = b2
..
.
am1 x1 + am2 x2 + · · · + amn xn = bm
Each of these linear equations gives a hyperplane. (In 2-space, a linear equation in x, y gives
a line. In 3-space, a linear equation in x, y, z gives a plane that we can draw.)

When we are in n-space, the intersection of n hyperplanes, generally speaking, gives us a


point; the intersection of n − 1 hyperplanes gives us an “edge”.

1
Candidates for the optimal solution of an LP are the corner points of the poly-
tope in n-space and they correspond to the intersection of n hyperplanes.

FACT: If an LP has an optimal solution, then if a corner point feasible (CPF) solution has
no adjacent CPF solution that gives a better objective value then the current CPF solution
is optimal.

Simplex Method Geometrically:

• Initialization Start with a corner point on the polytope.

• Optimality Test Will the value of z increase if we move to an adjacent CPF?

– If no, then this is the optimal solution x? so STOP.


– If yes, then continue.

• Determine the Direction of Movement Consider the edges along which the value
of z increases. Along which edge will z increase most rapidly?

• Determine where to Stop Slide along this edge until we reach a new hyperplane -
equality of one of the constraints. If we go any further we will violate a constraint. We
are now at a new CPF.

• Go back up to Optimality Test.

2
Simplex Method Algebraically:

Introduce slack variables w = (xn+1 , xn+2 , · · · , xn+m ) which take on the value of the
space between Ax and b. We define w = b − Ax. We then write the augmented form of
the LP:
maximize z = cT x
subject to Ax + w = b, x, w ≥ 0.
FACT: Any solution (x1 , · · · , xn+m ) to the system of equations

Ax + w = b (1)

with xi ≥ 0 is a feasible solution to the original LP.


An augmented solution is a solution for the original variables that has been augmented
by the corresponding values of the slack variables. That is, it is a solution of (1).

We can see that the faces/hyperplanes of the polytope come from the n + m equations
x1 = 0, x2 = 0, · · · , xn = 0, xn+1 = 0, · · · , xn+m = 0. That is, every face (hyperplane) of our
feasible region is associated with one of the n + m variables being equal to zero.

The fact that every corner of the feasible region is associated with the intersection of n
hyperplanes, i.e. simultaneous n variables are equal to 0, leads us to the realization that we
need to look at solutions to Ax + w = b where n of the n + m variables are zero and the
others are ≥ 0. These are called basic feasible solutions (BFS).

When you think of a corner point feasible (CPF) point, think simultaneously of the basic
feasible solution (BFS). They are the same except the CPF just reports the value of the
original variables x1 , · · · , xn whereas the BFS reports the slack values, too.

Of the n + m variables of a BFS, m of them are designated as basic variables and the other
n are non-basic variables.

The n non-basic variables are set to zero. Then the system Ax + w = b is solved to find the
values of the remaining m basic variables.

There are n + m variables and only m equations in (1) so when we force n of them to be
zero, we are left to solve the m equations for m unknowns which, generally but not always,
gives us a single solution.
As long as these basic variables that we solve for are ≥ 0, this is a BFS and it is associated
with a corner point on the feasible region.

You might also like