21–29 of 29 posts
Are those within the ballpark of what you need to do?
--
Am adding this response as both:
-1 Hey maybe you didn't know about this
-2 A question: Is CVXpy even in the same domain as those other tools listed here in the replies
I never understood why it is called linear 'programming'. Why programming instead of optimization ?
Earlier quoted context omitted.
I totally agree. But if you are a non-profit you might be able to get a SCIP license for free, you just have to contact the people behind it. But if a moderate performance improvement is not that important for you then GLPK is definetly the best option.
The problems we solve (and the timeframe we're under) are among the hardest in the OR sector. Not using CPLEX or GUROBI isn't an option. GUROBI has some benchmarks on their site that are illuminating.
I never understood why it is called linear 'programming'. Why programming instead of optimization ?
Googling "linear programming etymology" returns https://mathoverflow.net/questions/145077/why-are-optimizati...
Earlier quoted context omitted.
I work at a non-profit and not a University, so I can't use them for free. My very nice, but slightly used car cost me less money and it has been driven over an hour per day for 7 years. That makes those solvers very hard for me to justify outside of production where the license costs are worth every darn penny.
> I work at a non-profit and not a University, so I can't use them for free. What’s required to qualify for an academic license? Could you split your nonprofit into a business aspect, and then a walled-off research aspect (funded by the business aspect) that by itself fits the definition of academia by which these companies judge? Or, more simply, could your nonprofit partner with a University on R&D, with the Univer…
For a course at uni (am not a practitioner in this domain in any measurement) we used cvxpy (and there was cvxopt for the matlab inclined). Are those within the ballpark of what you need to do? -- Am adding this response as both: -1 Hey maybe you didn't know about this -2 A question: Is CVXpy even in the same domain as those other tools listed here in the replies
Also, I think cvxpy might implement an interior point method, not a simplex method, so you don't get a "vertex solution", which often has some nice sparsity attributes.
Earlier quoted context omitted.
> I work at a non-profit and not a University, so I can't use them for free. What’s required to qualify for an academic license? Could you split your nonprofit into a business aspect, and then a walled-off research aspect (funded by the business aspect) that by itself fits the definition of academia by which these companies judge? Or, more simply, could your nonprofit partner with a University on R&D, with the Univer…
Several of the solvers require that you have an academic email (or even IP) adress. If you don't but still think you qualify, I guess one would need to contact them and go through some work-around, possibly every year to renew?
Earlier quoted context omitted.
Bonus if you can use Subspace-BB (Barzilai-Borwein) on what is left after activating your constraints. That stuff is magical. Also because you can throw it at linear systems that are actually FFT-based time/space variant convolutions or similar things where you never, ever want so much as touch the sparse matrix of that linear system.
Y not touch the sparse matrix of that system?