max
Analysis · Concept hub

Linear Programming

1939 CEIndependent 20th-century routes through Soviet production planning and American military and administrative planning

Concept

Optimize a linear objective over a feasible region defined by linear constraints. Kantorovich’s 1939 production planning and Dantzig’s 1947 planning model and simplex method formed in different institutions; a model may be infeasible, unbounded, or have multiple optima.

Understand it in one breath

Optimize a linear objective over a region cut out by linear constraints. If a finite optimum exists and the feasible region has vertices, at least one optimal vertex exists. An entire edge can also be optimal, and a model may be infeasible or unbounded. Simplex and interior-point methods support resource allocation, flow, and production planning.

At a glance

Scenario

Constraints

Objective

Optimum

Factory resource allocation

Materials ≤ 100; labor ≤ 80

Profit = 5x + 3y → max

One of the vertices

Blending and mixing

Ingredients, nutrition, and demand

Cost → min

An LP when quantities are continuous

Network flow

Capacity and flow conservation

Cost → min

Some routing and transport problems

Linear approximation to power dispatch

Demand and capacity

Operating cost → min

Real systems add nonlinear and integer constraints

Some linear-classification variants

Linear constraints

Linear loss or cost → min

Can be an LP; a standard SVM is usually a quadratic program

The simplex algorithm moves among adjacent basic feasible solutions, while interior-point methods travel through the feasible region. Model quality also depends on how faithfully the objective and constraints represent reality.

Key formula

max  cxs.t. Axb,  x0\max\; \mathbf{c}^\top \mathbf{x} \quad \text{s.t. } A\mathbf{x} \leq \mathbf{b},\; \mathbf{x} \geq 0

Key moments

1939 CE

Kantorovich — production planning with scarce resources

In Leningrad, linear inequalities and solution methods were applied to production allocation. The work remained little known in the West for years and is not treated as a direct transmission into the later American route.

1947 CE

Dantzig — the planning model and simplex method

In US Air Force planning, Dantzig developed a method that improves an objective by moving among vertices of a feasible polytope. Programming meant organizing plans more than writing code.

1979 CE

Khachiyan — the first polynomial-time LP algorithm

The ellipsoid method showed that linear programming is solvable in polynomial time in theory. Worst-case complexity and practical speed are different comparisons.

1984 CE

Karmarkar — a polynomial-time route through the interior

Karmarkar opened a major era for interior-point methods, but his was not the first polynomial-time LP algorithm and it did not replace simplex in every setting.

Modern applications

Airline and logistics planning, network routing, portfolio optimization, and support-vector machines.

Beyond MathVoyage

Loading…