Linear Programs with Totally Unimodular Matrices updated 19 March 2009 Basic Feasible Solutions max s.t. z 5x 8 y x y 6 (1) 5 x 9 y 45 (2) x, y 0 Standard Form max z 5 x 8 y s.t. x y s1 5 x 9 y s2 x, y, s1 , s2 6 45 0 slide 1 Basic Feasible Solutions max z 5 x 8 y s.t. x y s1 5 x 9 y s2 x, y, s1 , s2 Solution Basic Variables 6 45 0 Non-Basic Variables Intersection BFS 1 x = 2.25 & y = 3.75 s1 = s2 = 0 (1) and (2) BFS 2 x = 6 & s2 = 15 y = s1 = 0 (1) and x-axis BFS 3 y = 5 & s1 = 1 x = s1 = 0 (2) and y-axis BFS 4 s1 = 6 & s2 = 45 x= y=0 x-axis and y-axis slide 2 Vector-Matrix Representation x y s1 s2 6 1 1 1 0 A b 45 5 9 0 1 B x, y, AB B x, s2 , AB B y, s1, AB B s1 , s2 , AB 1 5 1 5 1 9 1 0 1 , 9 0 , 1 1 , 0 0 , 1 1 B A b AB1b AB1b AB1b 2.25 3.75 6 15 5 1 6 45 slide 3 Example MCNFP -2 (3, 2,5) 5 1 2 (4, 1,3) 4 (1, 0,2) (2, 0,2) 3 -3 (4, 0,3) 0 slide 4 LP for Example MCNFP Min 3x12 + 2 x13 + x23 + 4 x24 + 4 x34 s.t. x12 + x13 = 5 {Node 1} x23 + x24 – x12 = -2 {Node 2} x34 – x13 - x23 = 0 {Node 3} – x24 - x34 = -3 {Node 4} 2 x12 5, 0 x13 2, 0 x23 2, 1 x24 3, 0 x34 3, slide 5 Matrix Representation of Flow Balance Constraints 1 1 0 0 1 0 0 0 0 1 1 0 1 1 0 1 0 0 1 1 x 5 12 x13 2 x23 0 x 24 3 x34 slide 6 Solving for a Basic Feasible Solution 1 1 0 0 1 0 0 x 0 1 0 x 1 0 1 x 0 1 1 x x 12 x13 x24 x 34 5 2 13 0 24 3 34 12 1 1 0 1 0 1 0 1 0 0 0 1 1 0 5 0 2 0 1 1 3 slide 7 Cramer’s Rule Use determinants to solve x=A-1b. x j B A j aij , B j an1 ann a b a a b 12 n2 1 n 1n Take the matrix A and replace column j with the vector b to form matrix Bj. slide 8 Using Cramer’s Rule to Solve for x12 x 12 5 1 0 0 2 0 1 0 0 1 0 1 3 1 0 1 1 1 0 0 1 0 1 0 0 1 0 1 0 Is B(1, 2 ) Does B(1, 2 ) A 0 1 1 an integer? A 1? slide 9 Total Unimodularity • A square, integer matrix is unimodular if its determinant is 1 or -1. • An integer matrix A is called totally unimodular (TU) if every square, nonsingular submatrix of A is unimodular. 1 1 1 1 Not TU 1 1 0 0 1 0 1 0 0 1 0 1 0 0 1 1 TU slide 10 Total Unimodularity • A square, integer matrix is unimodular if its determinant is 1 or -1. • An integer matrix A is called totally unimodular (TU) if every square, nonsingular submatrix of A is unimodular. 1 1 0 0 1 0 1 0 0 1 0 1 0 0 1 1 1 0 1 1 1 slide 11 Sufficient Conditions for TU An integer matrix A is TU if 1. All entries are -1, 0 or 1 2. At most two non-zero entries appear in any column 3. The rows of A can be partitioned into two disjoint sets M1 and M2 such that • • If a column has two entries of the same sign, their rows are in different sets. If a column has two entries of different signs, their rows are in the same set. slide 12 The Matrix of Flow Balance Constraints x 12 1 5 1 0 0 0 x13 0 1 1 0 2 1 x 23 0 1 1 0 0 1 x 24 0 3 0 0 1 1 x34 • Every column has exactly one +1 and exactly one -1. • Conditions 1 and 2 are satisfied. • Let the row partition be M1 = {all rows} and M2 = {}. • Condition 3 is satisfied. • Therefore, the flow balance constraint matrix is TU. slide 13 Using Cramer’s Rule to Solve for x12 x 12 5 1 0 0 2 0 1 0 0 1 0 1 3 1 0 1 1 1 0 0 1 0 1 0 0 1 0 1 0 Is B(1, 2 ) an integer? Does A 1? Yes. B(1, 2 ) A 0 1 1 slide 14 Expansion by Minors: 4-by-4 Matrix a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a 11 21 31 41 12 22 32 42 22 11 23 24 12 31 13 14 13 23 33 43 32 33 34 22 23 24 14 24 11 34 43 14 44 42 43 44 12 13 44 42 a a a a a a a a a a a a a a a a 21 22 23 24 12 21 13 14 12 41 13 14 a a a a 31 32 33 34 a a a a a a 32 33 34 22 23 24 a a a a a a a a a a 41 42 43 44 42 43 44 32 33 34 slide 15 Expansion by Minors: 3-by-3 Matrix a a a a a a 11 12 21 22 31 32 a a a 22 11 32 a a a 13 23 33 a a 23 a12 a 21 a 33 31 a a 23 33 a13 a 21 a 31 a a 22 32 a a a a a 11 22 33 23 32 a a a a a 12 21 33 23 31 a a a a a 13 21 32 22 31 slide 16 Using Cramer’s Rule to Solve for x12 x 12 5 1 0 0 2 0 1 0 0 1 0 1 3 1 0 1 1 1 0 0 1 0 1 0 0 1 0 1 0 Is B(1, 2 ) Does B(1, 2 ) A 0 1 1 an integer? A 1? Yes. slide 17 Using Cramer’s Rule to Solve for x12 5 1 0 0 5 2 0 3 2 0 1 0 1 0 1 0 0 1 0 1 0 1 0 1 3 0 1 1 0 0 1 1 • When we expand along minors, the determinants of the submatrices will be +1, -1, or 0. • Therefore, the determinant will be an integer: (5)(+1, -1, or 0) + (-2) (+1, -1, or 0) + 0 + (-3) (+1, -1, or 0). slide 18 Using Cramer’s Rule to Solve for x12 x 12 5 1 0 0 2 0 1 0 0 1 0 1 3 1 0 1 1 1 0 0 1 0 1 0 0 1 0 1 0 Is B(1, 2 ) Does B(1, 2 ) A 0 1 1 an integer? Yes. A 1? Yes. slide 19 TU Theorems • • Matrix A is TU if and only if AT is TU. Matrix A is TU if and only if [A, I] is TU. – • • I is the identity matrix. If the constraint matrix for an IP is TU, then its LP relaxation has an integral optimal solution. The BFSs of an MCNF LP are integer valued. slide 20