-------- Forwarded Message -------- From: Michal Trněný <[email protected]> To: [email protected] Subject: wrong answer Date: Thu, 06 Aug 2015 14:18:33 +0200 (CEST)
Here is a simple LP problem that GLPK didn't solve correctly: maximize objective: 4*x1 + x2 + 2*x2 given constraints: var x1>=0; var x2>=0; var x3>=0; c1: x1 <= 1; c2: x2 <= 1; c3: x3 <= 1; c4: x1 + x2 + x3 = 2.5; Solver yielded answer: x1=1, x2=1, x3=0.5, obj=7 that is obviously wrong, because: 4*1 + 1 + 2*0.5 != 7 but 4*1 + 1 + 2*0.5 = 6 and this solution (1,1,0.5) isn't optimal, because ... there exist solution (1,0.5,1), which objective is 6.5 - bigger, than objective of (1,1,0.5). You may imagine feasible region as unit cube with vertex (1,1,1) cuted off. I'd like you to nottice me, when you fix this bug. I want commented pseudocode of GLPK. Do you have it? Thank you. -- Michal Trněný www.michaltrneny.cz _______________________________________________ Bug-glpk mailing list [email protected] https://lists.gnu.org/mailman/listinfo/bug-glpk
