Extreme points

Extreme Points

An extreme point is a feasible point that cannot be written as a nontrivial convex combination of two different feasible points.

Definition

A point $x\in P$ is extreme if there do not exist distinct $y,z\in P$ and $0<\lambda<1$ such that

\[x=\lambda y+(1-\lambda)z.\]

Meaning

An extreme point is not in the middle of a feasible line segment. In a polyhedron, this is the same idea as a vertex.

Example

In the interval $[0,1]$, the points $0$ and $1$ are extreme. The point $1/2$ is not, because

\[\frac12=\frac12(0)+\frac12(1).\]

LP Importance

If $x=\lambda y+(1-\lambda)z$, then

\[c^Tx=\lambda c^Ty+(1-\lambda)c^Tz.\]

So an interior point of a segment cannot be strictly better than both endpoints. This is why LP optima can be searched at extreme points.

Equivalence

For polyhedra:

\[\text{extreme point}=\text{vertex}.\]

In standard form, these correspond to basic feasible solutions.

Checklist

You should be able to use the convex-combination definition and explain why endpoints matter in linear optimization.

See Also

Exam checkpoint

For BFS questions, convert to standard form before checking the point. A candidate in original variables must be lifted with slack and surplus variables before testing basis columns.

25

25
Ready to start
Extreme points
Session: 1 | Break: Short
Today: 0 sessions
Total: 0 sessions