Algoritmo Dual Simplex | Physics & Mathematics | Mathematics

7.

1

Dual Simplex Algorithm
Katta G. Murty, IOE 510, LP, U. Of Michigan, Ann Arbor, Winter 1997.

Example: Consider the following LPs:
x1 x2 x3 1 0 0 0 0 1 0 0 x4 x5 x6 − z 2 0 0 0 b −6 −2 4

0 −1 −2 0 1 0

1 −1 −2 1 3 1 −1 8 10

1 −100

xj ≥ 0 ∀j , min z x1 x2 x3 1 0 0 0 0 1 0 0 x4 x5 x6 − z 2 0 1 0 0 0 b −6 −2 4

0 −1 −2 0 1 0 1 2

1 −1 3

8 10

1 −100

xj ≥ 0 ∀j , min z
150

Dual simplex algorithm is just the opposite of the primal simplex algo.e. a dual feasible basic vector. Starting with a dual feasible basis (i. and let inverse tableau be: 151 .. INITIAL SETUP: Compute the inverse tableau corresponding to the initial dual feasible basic vector. GENERAL ITERATION: Let xB be present dual feasible basic vector. All operations are carried out on the primal simplex tableaus themselves. one in which c ¯j ≥ 0 for all j ) it tries to attain primal feasibility while maintaining dual feasibility throughout. subject to where Am×n and rank m.Consider LP in standard form: Ax = b. x ≥ 0 min z = cx. The Algorithm INPUTS NEEDED: LP in standard form.

xB opt. For each i ∈ F . Select one of these rows. . roe i eligible to be choosen as pivot row. PIVOT ROW SELECTION: F = {i : ¯ bi < 0}. the nonbasic relative costs c ¯j are all ≥ 0. 152 . ¯ bm −z −π ¯ 1 −z 0 All nonbasics = 0 Since xB dual feasible. sol. OPTIMALITY CRITERION: If ¯ bi ≥ 0 for all i = 1 to m. xB B −1 0 ¯ bi . . . the present rth basic variable is the dropping basic variable in this step.Inverse tableau wrt xB Basic Inverse tableau Basic var. values ¯ b1 . . and present BFS an opt.. say r ∈ F as pivot row. terminate. basic vector. So.

a ¯rn) ≥ 0. SELECTION: THE DUAL SIMPLEX MIN RATIO TEST: The entering variable (and hence the pivot col.(A. . a ¯rn) ≥ 0. and a ¯rj < 0} 153 . Find δ = min{−c ¯j /a ¯rj : xj nonbasic. PRIMAL INFEASIBILITY CRITERION: If pivot row = (¯ ar 1 . . the equation corresponding to it is: a ¯ r 1 x1 + .e. . 2. it cannot have a nonnegative solution. . . and go to infeasibility analysis. . (A . .) is selected to make sure that the next basic vector is also dual feasible. cost coeffs. . Terminate algo. DUAL SIMPLEX MIN RATIO TEST: If (¯ ar 1 . and coefficients of all variables ≥ 0. a ¯rn. i. . . . + a ¯rnxn = ¯ br with RHS constant ¯ br < 0.b) = (¯ ar 1 . which will be its updated col. let c ¯j denote rel. . So. Involves following work: 1.PIVOT COL.¯ br ) = (B −1)r. . . . . primal system infeasible. 3.¯ br ) . . Compute pivot row = updated rth row = ¯r. .. .

and let j = s be any of the subscripts that attains min above. Choose xs as entering variable. 4. Examples: Original tableau x1 x2 x3 1 0 0 0 0 1 0 0 0 x4 x5 x6 − z 7 0 b 8 4 −5 0 −2 1 0 4 −2 2 2 0 −2 0 1 2 0 1 −3 1 3 xj ≥ 0 for all j . min z 154 . Perform pivot step and go to next iteration with new basic vector.

Original tableau x1 1 x2 1 x3 x4 x5 x6 − z 1 1 0 0 0 0 1 0 0 0 0 1 0 0 b 1 −1 −2 −1 1 8 3 24 2 15 0 −8 0 1 5 0 xj ≥ 0 for all j . min z 155 .

Let K denote the set of feasible solutions for this original problem. subject to Ax = b. So. larger than 156 . Opt. an optimum sol.6 Properties.7. x ≥ 0. 1. So. we verify that x x ¯∈K ¯ is in fact ˜ . of relaxed problem xB . i. Let K obtained by relaxing the constraints xj ≥ 0 for each j ∈ J ˜ ⊃ K and violated by x ¯. is an opt. from the original problem.. sol. Each Sol. with the set of feasible sols. in the dual simplex algo. Let xB be a dual feasible but primal infeasible basic vector associated with basic sol. every basic sol. for a relaxed problem. for the a minmum cost solution in K relaxed problem mentioned above. x ¯ violates the primal constraints xj ≥ 0 for j ∈ J ..e. Let J = {j : x ¯ j < 0} . in a larger set: Consider the problem in standard form: min z = cx. x ¯. K ˜ . and each of these variables xj for j ∈ J is a basic variable in ˜ denote the set of feasible sols. Since xB is dual feasible.

this variable is nonbasic. do not change. In a degenerate pivot step. value is approached from below. value and dual sol. in dual simplex algo. but primal sol. changes strictly in every pivot step. value of current basic sol. 2. value strictly increases. value in original problem. opt. the obj. in dual simplex algo. pivot step is: nondegenerate degenerate if minratio is > 0 if min ratio is 0. obj. 157 . the obj. is always a lower bound for min obj. in a monotonic increasing manner. 3. So obj. also satisfies the bound restriction on entering variable. The dropping variable is always one the bound restriction on which is violated by present sol. so the bound restriction for it holds as an equation in the next basic sol. So. In next basic vec. This sol. as it has positive value in it. As before.K . In a nondegenerate pivot step.

very useful in sensitivity analysis.7. if changes occur in RHS constants vector. After problem solved. Dual Simplex Method If an initial dual feasible basis not available. this method not used much in practice. an artificial dual feasible basis can be constructed by getting an arbitrary basis. 158 . However. and then adding one artificial constraint. dual simplex algo. because the dual simplex min ratio test needs O(n) comparisions in every pivot step (primal simplex min ratio test needs only O(m) comparisons in each step. and in most real world models n >> m).8 Usefulness of Dual Simplex Algorithm Not used to solve new LPs. Though mathematically well specified. dual simplex iterations are used to get new opt.

Sign up to vote on this title
UsefulNot useful