[Scip] ENFOLP vs. SEPALP

[Daniel Karch]

Hi,

I have a few questions about the SEPALP and ENFOLP callbacks of a
constraint handler.
I am still not completely sure when to use what ...

1. As far as I understand, SEPALP should add cutting planes to the LP while
it is being solved,
and ENFOLP uses cutting planes to cut off an optimal LP solution. Is that
correct?
Does that mean that SEPALP also works on non-optimal LP solutions?

2. When a cutting plane is added via SCIPaddCut, which nodes "know" about
it? Suppose
node 0 creates a branching with children 1 and 2, and node 1 is treated
before node 2. If
I add a cutting plane in the SEPALP or ENFOLP callback of node 2 which is
violated also in
node 3, will it be separated automatically, or will it be violated again in
node2? I.e., do I have
to add it as a pool cut if I want it to be separated in node 2?

3. In the TSP example, the following lines in the ENFOLP callback

    // if a subtour was found, we generate a cut constraint saying that
there must be at least two outgoing edges
    if( found )
      *result = SCIP_INFEASIBLE;

suggest that you do not generate a cut after all, but return
SCIP_INFEASIBLE instead. Why?


Best,

  Daniel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listserv.zib.de/mailman/private/scip/attachments/20130416/d87bd99d/attachment.html