There is no single best solver for all "complex optimization problems of several tens of thousands of variables and several thousand constraints". So this question cannot really be answered. In general, however, one would look at commercial solvers such as Cplex (but distributed mip is deprecated) and Gurobi. I think most modelers use a reasonable SMP machine (just easier and they can easily be equipped with enough cores and memory that distributed MIP is not worthwhile).
Note that a model with 10k variables and equations is not very large these days. The model may still be difficult, and a lot depends on the skill of the modelers.
发布评论
评论（1）
对于所有“数十万个变量和数千个约束的复杂优化问题的复杂优化问题”，没有一个最佳求解器。因此，这个问题无法真正回答。但是，通常，人们会考虑商业求解器，例如CPLEX（但分布式MIP被弃用）和Gurobi。我认为大多数建模者都使用合理的SMP机器（只是更容易，并且很容易配备足够的内核和内存，而分布式MIP不值得）。
请注意，如今具有10K变量和方程式的模型并不是很大。该模型仍然很困难，很大程度上取决于建模者的技能。
There is no single best solver for all "complex optimization problems of several tens of thousands of variables and several thousand constraints". So this question cannot really be answered. In general, however, one would look at commercial solvers such as Cplex (but distributed mip is deprecated) and Gurobi. I think most modelers use a reasonable SMP machine (just easier and they can easily be equipped with enough cores and memory that distributed MIP is not worthwhile).
Note that a model with 10k variables and equations is not very large these days. The model may still be difficult, and a lot depends on the skill of the modelers.