A bounded linear objective reaches its last feasible contour at a corner. This lesson compares the flagship objective, a tilted objective with a different winner, and an honest edge tie. The diagram shows how the same vertex rule handles all three.

highlighted = computed this step

Last touch

Slide the contour outward until it last touches the feasible region. The touching corner is (4/3, 4/3). Why: the best value is the final contour that still has any feasible contact.

last touch at (4/3,4/3)\text{last touch at }(4/3, 4/3)
Last-touch contoursEach recomputed optimum is the point or edge touched by its final contour.Feasible region2x + y ≤ 4x + 2y ≤ 4z=8/3(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=8/3Feasible region2x + y ≤ 4x + 2y ≤ 4z=6(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=6Feasible regionx + y ≤ 2x ≤ 3y ≤ 3z=2(0, 0)(0, 2)(2, 0)multiple optima, z=2

Objective value

At that corner, z=x+y=4/3 plus 4/3=8/3. Why: the objective value is recomputed from the same exact point the dot marks.

z=x+y=4/3+4/3=8/3z=x+y=4/3+{}4/3=8/3
Last-touch contoursEach recomputed optimum is the point or edge touched by its final contour.Feasible region2x + y ≤ 4x + 2y ≤ 4z=8/3(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=8/3Feasible region2x + y ≤ 4x + 2y ≤ 4z=6(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=6Feasible regionx + y ≤ 2x ≤ 3y ≤ 3z=2(0, 0)(0, 2)(2, 0)multiple optima, z=2

Different winner

With z=3x+y on the same polygon, the recomputed winner moves to (2, 0) with z=6. Why: changing the objective direction can change which corner is touched last.

max(3x+y)=6 at (2,0)\max(3x+y)=6\text{ at }(2, 0)
Last-touch contoursEach recomputed optimum is the point or edge touched by its final contour.Feasible region2x + y ≤ 4x + 2y ≤ 4z=8/3(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=8/3Feasible region2x + y ≤ 4x + 2y ≤ 4z=6(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=6Feasible regionx + y ≤ 2x ≤ 3y ≤ 3z=2(0, 0)(0, 2)(2, 0)multiple optima, z=2

Whole edge tie

For the tie example, a whole edge is optimal: the endpoints (0, 2) and (2, 0) both have z=2. Why: when the final contour lies on a boundary edge, the optimum is not a single point.

maxz=2 on an edge\max z=2\text{ on an edge}
Last-touch contoursEach recomputed optimum is the point or edge touched by its final contour.Feasible region2x + y ≤ 4x + 2y ≤ 4z=8/3(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=8/3Feasible region2x + y ≤ 4x + 2y ≤ 4z=6(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=6Feasible regionx + y ≤ 2x ≤ 3y ≤ 3z=2(0, 0)(0, 2)(2, 0)multiple optima, z=2

Vertex rule

For a bounded linear program, a maximum occurs at a vertex, so checking corners is enough. The recomputed optimum is (4/3, 4/3) with z=8/3. Why: even an edge tie is detected from its endpoint vertices.

maxz=8/3 at (4/3,4/3)\max z=8/3\text{ at }(4/3, 4/3)
Last-touch contoursEach recomputed optimum is the point or edge touched by its final contour.Feasible region2x + y ≤ 4x + 2y ≤ 4z=8/3(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=8/3Feasible region2x + y ≤ 4x + 2y ≤ 4z=6(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=6Feasible regionx + y ≤ 2x ≤ 3y ≤ 3z=2(0, 0)(0, 2)(2, 0)multiple optima, z=2

Diagram note

The last-touch contour may name one corner, a different corner, or a whole edge. Pixel positions are rounded for layout; every number shown is exact.

last touch identifies the optimum set\text{last touch identifies the optimum set}
Last-touch contoursEach recomputed optimum is the point or edge touched by its final contour.Feasible region2x + y ≤ 4x + 2y ≤ 4z=8/3(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=8/3Feasible region2x + y ≤ 4x + 2y ≤ 4z=6(0, 0)(0, 2)(4/3, 4/3)(2, 0)optimal z=6Feasible regionx + y ≤ 2x ≤ 3y ≤ 3z=2(0, 0)(0, 2)(2, 0)multiple optima, z=2