I have extracted some of the best solutions that Gurobi identified in the solution pool. Surprisingly, I have noticed that Gurobi is indeed capable of producing the solution I desire. However, Gurobi does not recognize this preferred solution as the best one in its evaluation. The solution I prefer can be confirmed as optimal using conventional methods and is much closer to the known actual value. After running thousands of cases, I have encountered this scenario several times. I am confused about why it cannot always find the real optimal solution for my model.
I am working on an MIQP problem with an objective function of the form v'Wv, where v is a vector containing both continuous and integer parameters that need to be estimated, and W is a symmetric positive definite matrix.
Please sign in to leave a comment.