Forum Discussion
FICO
8 months agoCommunity Manager
Implementation of B&B with python
HI! I'm trying to implement a branc-and-bound solver by using the python interface of Xpress. I implemented a MINLP model, and I am trying to solve the linearization with the B&B algorithm. Even t...
FICO
8 months agoCommunity Manager
Expanding on Imre's answer:
- There is no explicit way to track the cutoff. However, the cutoff should only be updated from new feasible solution, so you could look at the CURRMIPCUTOFF attribute from the intsol callback.Or you could use the cutoff passed into the preintsol callback. You could also use the gapnotify callback, that notifies about changes in the gap.
- Are you sure you processed more than one node?
- There is no way to print a list of active nodes. But you can follow the B&B process by means of different callbacks and attributes. For example, you can use a prenode callbackto get notified about new node being started. At that point it the branchvar and branchvalue should tell you the branching decision that led to this node.Or use the newnode callback to get notified about any node being created (which may be way before it is actually processed).
- Could you use the newnode callback to double check which nodes are actually created? If this does not help, can you please share your code with danieljunglas(at)fico(dot)com?
- In the change branch object callback there is no branching decision made yet. The callback only tells which branches Xpress should create, not which branch to follow next. So the branchvalue and branchvar attributes are meaningless in this context.
Related Content
Related Content
- 0Views0likes1Comment
How can I use ODH for solving a MIP model using Xpress Python API?
Hi everyone, I want to play a bit with ODH for Xpress Solver. I did run ODH from the command line but now I want to integrate it in the whole pipeline of a Python program. What should I import? What...1 month agoPlace FICO® Xpress Optimization ForumFICO® Xpress Optimization Forum2 unread comments2 unread4Views0likes2Comments