Recently I’ve been using a really great linear programming / mixed integer programming package called Gurobi. As a combinatorialist I use it almost exclusively for integer programming, which is basically mixed integer programming where the “mixture” is 100% – 0% integer to continuous variables.
It’s a full commercial product, which usually would cost tens of thousands of dollars, but – unusually – they offer a completely free academic license for the full-scale product (i.e. not some size-limited or cpu-time-limited version) which is a trend that I would heartily like to encourage, particularly as academic funding (in Australia) continues its inexorable year-on-year decline. In principle, it comes with no support which is fair enough, but in practice there is a forum that is closely monitored by a couple of guys who seem to respond quickly and courteously to pretty much any query. I found and reported a bug which they seemed pleased about, and so I feel that in some small way I’m contributing back to them.
But what a funny name – why would you call anything “Gurobi”?
Let’s move to a seemingly unrelated topic – the Maastricht workshop on matroids that I posted about two weeks ago. One of the participants was a mathematician called Bob Bixby who, many years ago, characterized the excluded minors for ternary matroids (that is, representable over the finite field ). At first sight there’s no obvious relationship between the arcane area of pure mathematics known as matroid theory and hard-core industrial optimization, but in fact there are fundamental connections. Bixby was at the forefront of both areas at this time and was one of the original founders and developers of the incredibly influential linear programming package CPLEX, now owned by IBM.
As it happens, Gurobi is Bob Bixby’s latest venture in commercial optimization packages – well, not Bixby alone, but a trio of researchers called GU, ROthberg and BIxby – and hence the unusual name.
I just wish I’d made this connection two weeks ago and taken the opportunity to thank Bob personally for the academic license policy!