Originally posted by: moqua

Hey,

I'm a beginner in ILOG and I'm still desperatly trying to find information whether ILOG is the right tool for me at all.
I'm doing a research on architecture exploration/optimization in the context of embedded systems. So I basically look at tool, where one can model a system architecture and which provides some kind of methods in optimizing this architecture. Does ILOG something like that at all? As far I understand its mostly about some buisness processes or something?

-Sergey
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

IBM ILOG CPLEX solves so-called "mixed integer programs", which are of the form

min c*x
s.t. A*x <= b
     x >= 0
     x_j integer for j \in I

You can also have quadratic terms in the objective and/or the constraints, as long as the problem stays convex. Now, the big question in your case is: can you model your optimization problem using linear or quadratic inequalities and equations and continuous and integer variables.

You should probably read an introductory book (or take a class) in mathematical programming and in particular modeling. Modeling a problem as mixed integer program is a non-trivial task, but once you have done it CPLEX is a very powerful tool to solve and analyze your problem.

Tobias
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Extending Tobias's answer, IBM's ILOG product series (I'm not sure what the official phrasing for that is) includes another solver (CPOptimizer) and an modeling language (OPL) that works with both CPLEX and CPOptimizer. CPOptimizer is designed to solve constraint programming (CP) problems. CP problems are more general than linear, quadratic and mixed-integer linear programs in that they do not limit you to linear (or at most quadratic) functions, do not require convexity, and support more general constraints than what CPLEX can handle. The downside is that CP models are really best suited for working with variables with finite (preferably relatively small) domains. Continuous variables are problematic for them. So, depending on what your models would look like, you might want to use CPLEX, you might want to use CPOptimizer, or you might want to look elsewhere.

Paul

Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers
#DecisionOptimization