[Scip] Branch on original variables

[Gerald Gamrath]

Dear Bahareh,

>> Second, I am giving an initial feasible solution to the master problem
>> when creating it, and the corresponding dual variables are zero (Those
>> duals that are obj coefficient in the subproblems). These cause my
>> subproblems to add zero columns to the master problem. do you think
>> there is a way to fix this?or I just have to change the initial 
>> solution?
>
> I am not sure I understand your question.  But maybe the 
> branch-and-price experts can answer this.

I'm also not sure I understand what you mean. Why is it a problem that 
zero columns are generated? If they are feasible and help to improve 
your LP solution, this is what should happen (and then, perhaps your 
initial solution could be improved).

To get some more details:
All dual solution values are zero? But the reduced cost of the created 
zero columns are negative? This means that their objective coefficient 
is negative?

Perhaps you missed some dual variables? Do you have bounds on variables? 
They also lead to a variable in the dual problem.

Also, can you check the LP solution status via SCIPgetLPSolstat(). It 
should be SCIP_LPSOLSTAT_OPTIMAL.

Best,
Gerald