Branch & Price

From: Jesper Hansen Ph.D.Student (jc 1/2002) <jha_at_imm.dtu.dk>
Date: Wed 05 Feb 2003 08:56:54 AM GMT
Message-ID: <3E40D1D6.6080300@imm.dtu.dk>
Hi

I'm doing a Branch & Price algorithm. Instead of solving the primal 
master problem, I'm solving the dual master problem, adding rows instead 
of columns. In this way, I can simply post the generated columns as rows 
and resolve instead of setting up the entire problem at every resolve.

Now, to do branching, I need to access the primal variables or when 
solving the dual problem the values of the dual variables, in order to 
check if the solution is integral and if not find constraints to branch 
on. I notice that fewer dual values are returned by eplex than added 
rows. This is unfortunate since I don't know which, and what the values 
are of the remaining. A way around this is to setup and solve the primal 
master problem in each master iteration, but again I cannot be sure to 
get all the dual variables to send to the pricing problem. I hope the 
conclusion is NOT that I'll have to solve both the primal and the dual 
problem in each iteration, which seems a bit inefficient and cumbersome. 
Alternatively, if I could be sure that all posted rows were actually 
sent to the solver without any reductions from ECLiPSe, then I would 
receive all dual variables of the dual master problem (the primal 
variables of the primal master problem). Is that possible or do you have 
any other suggestions?

'Jesper

-- 
_______________________________________
Jesper Hansen         
Ph.D. student         
Telephone: (+45) 45 25 33 88 
Telefax.:  (+45) 45 25 26 73
E-mail: mailto:jha@imm.dtu.dk 
Homepage: http://www.imm.dtu.dk/~jha/
Department of Mathematical Modelling 
Building 305                      
Technical University of Denmark
DK-2800 Lyngby 
Received on Wed Feb 05 09:00:08 2003

This archive was generated by hypermail 2.1.8 : Wed 16 Nov 2005 06:07:21 PM GMT GMT