I am actually constructing engineer application based on pseudoboolean optimization. I want to ask what is the current status (how many variables, interaction parameters) the solver generally solve up to date and what is the best software for it. Thank you. asked 21 Oct '14, 19:55 Chivalry 
When you say pseudoboolean optimization, do you mean your objective function looks like this? My first approach would be to linearize and use a commercial MIP solver (disclaimer: I have no experience solving pseudoboolean optimization problems). You might look at this survey from 2002. Also look at which papers cite it for more recent findings and computational results. As far as what size problems can be solved, my guess is that YMMV depending on what instances you're looking at. I bet I can find 500variable instances that cannot be solved by current techniques, but these instances are artificial and probably not what you'd encounter in practice. answered 21 Oct '14, 23:16 Austin Buchanan I did the same thing using MIP.... maybe due to lack of knowledge, instances with 30~40 variables is already rather slow....
(21 Oct '14, 23:35)
Chivalry
hmmm, that's pretty badworse than I expected. How big are the multlinear terms? And how many variables (binary vs. continuous) do you have in the linearization?
(21 Oct '14, 23:51)
Austin Buchanan
