[Coin-discuss] Benchmark Instances & code for format conversions

[Antonio Frangioni]

I understand your doldrums. I think everybody who got a Ph.D. in  
computational optimization
went through your paces. That's why ...

> PS2: I am searching for benchmark instances (both structured and  
> unstructured MIPs) and would appreciate any pointers to publicly  
> available instances besides the following: MIPLIB, Mittleman  
> Instances, CORAL Instances, OR-Library, Atamturk Instances, DIMACS  
> challenge instances.

... I've struggled to maintain a library of instances. Unfortunately  
the .mps format is
not the one of my choice, although I've developed tools that (among  
other things) can
convert *some* of the formats into .mps.

My library is about Multicommodity-like problems. You can find it at  
my web page

	www.di.unipi.it/~frangio

(link "data"). It's mostly about continuous problems, but there are a  
few groups of
instances (and links) that may interest you.

					Hope this helps

					   Antonio

P.S. I believe that the only way to overcome the problem you're  
lamenting is to
      put up, collect everything you can and be the first to make it  
available,
      so as to be you to be the "seed" of *the great MPS Internet  
collection*, or
      whatever.

P.P.S. One serious issue is that of the format. MPS is ugly, and it's  
especially
        bad at *hiding* any existing, clean structure in your model  
in a way that
        makes it very difficult to recover it. That's why, given that  
I work on
        *structured* problems, I don't support MPS. There have been  
some attempts
        at overcoming this: I sincerely hope that they will become  
mainstream as
        soon as possibly, finally relegating MPS to the dusty shelf  
full of punched
        cards where it rightly belongs ...