When solving a MAX problem and a variable is pushed to its LOWER bound cbc is reporting a positive value for the reduced cost. I believe in practice it should be reported as negative. For the following three complementary cases, cbc seems to report the correct sign for the reduced cost
max problem -> var at UB -> (+) r.c. min problem -> var at LB -> (+) r.c. min problem -> var at UB -> (-) r.c.
The problematic case (which occurs when using the LP file interface) can be tested using the following simple LP file:
max
obj:
-1.000000 x(1)
s.t.
bounds
-1 <= x(1) <= 1
end
This happens in 2.7, 2.8, and trunk. Below is my contact info:
name: Gabe email: gabehack@…
=================================
I don't like how some of that formatting came out. See below for what are the working cases...
When solving a MAX problem and a variable is pushed to its LOWER bound cbc is reporting a positive value for the reduced cost. I believe in practice it should be reported as negative. For the following three complementary cases, cbc seems to report the correct sign for the reduced cost
max problem -> var at UB -> (+) r.c. min problem -> var at LB -> (+) r.c. min problem -> var at UB -> (-) r.c.
The problematic case (which occurs when using the LP file interface) can be tested using the following simple LP file:
This happens in 2.7, 2.8, and trunk. Below is my contact info:
name: Gabe email: gabehack@…
=================================
I don't like how some of that formatting came out. See below for what are the working cases...
max problem -> var at UB -> (+) r.c.
min problem -> var at LB -> (+) r.c.
min problem -> var at UB -> (-) r.c.