Originally posted by: J5JK_chao_lei

I was trying to pass the basis information of last program to the next by calling writeBasis and readBasis. But the warning message "Slack removed by singularity." displayed in the output result. This program seems running correctly even with the presence of the warning.

Could someone tell me what does this information indicate? Should I be worried about it? Thanks a lot!
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: RWunderling

CPLEX has an auto-repair feature for singular bases that repairs the singularity by replacing columns with slacks in the basis. However, when a slack itself is involved in a singularity such a replacement is not very useful, which is what this warning is about. Nevertheless, CPLEX can recover from it, but your performance may not be better or even worse than when solving your model from scratch.

Thus, you should really question your use of your previous basis, i.e. make sure that the changes of the model you perform are consistent with keeping the basis non-singular, or better maintaining primal or dual feasibility. Consider for instance the following LP:

c1: x + y <= b
c2: x <= c

If you now install a basis declaring y and the slack of c1 basic, you end up with the following singular basis matrix:

(1 1)
(0 0)

which could trigger the warning message you are seeing. Installing such a basis (via readBasis or directly) will likely not help speed up the solution of your model. If your algorithm dictates such changes, you should consider turning off the advance indicator to instruct CPLEX to solve the modified problem from scratch.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: J5JK_chao_lei

Thanks a lot for your answer!
#CPLEXOptimizers
#DecisionOptimization