LocalSolver is a new kind of mathematical optimization solver. It combines different optimization technologies to solve your problem at hand. Using LocalSolver, you can now tackle ultra-large, real-life, discrete, numerical, or even black-box problems in a model-and-run fashion without any tuning. Because efficiency is nothing without reliability and robustness, we ensure our clients the best quality of products and services thanks to a drastic continuous integration methodology coupled with a responsive and dedicated support.
All-terrain & all-in-one optimization solver
LocalSolver combines many different optimization techniques while needing no parameter tuning: local and direct search, constraint propagation and inference, linear and mixed-integer programming, as well as nonlinear programming techniques.
Contrarily to other math optimization software, LocalSolver is not based on a single optimization technique. It hybridizes different optimization techniques dynamically, during the resolution, thanks to a unique search approach. LocalSolver combines local and direct search techniques, constraint propagation and inference techniques, linear and mixed-integer programming techniques, as well as nonlinear programming techniques, to solve your problem at best.
Moreover, LocalSolver integrates unique heuristic search techniques for both combinatorial and continuous optimization. In this way, it is able to tackle models involving millions of variables, which are out of scope of classical solvers, especially classical mixed-integer linear programming (MIP), constraint programming (CP), or nonlinear programming (NLP) solvers.
Scalability of OR solution technologies
Innovative math modeling language
LocalSolver comes with a powerful modeling language. It enables you to quickly prototype your optimization applications.
The LocalSolver Programming language (LSP) offers an efficient programming style: dynamic but strongly typing, implicit variable declaration, compact looping syntax, etc. Many functions can be used both for mathematical modeling or for programming, making the language easy to learn.
Our goal is to reduce your programming effort as much as possible (efficiency), while framing your prototyping work (reliability). You will see that the resulting LSP models are less verbose and more readable than the ones written with existing modeling languages.
A piece of LSP model
Lightweight object-oriented APIs
To fully integrate LocalSolver in your business applications, we provide easy-to-use object-oriented programming interfaces for Python, Java, C#, C++.
LocalSolver's APIs are lightweight, exposing only a few classes. The corresponding callable libraries (Python, Java, C#, C++) are provided common operating systems (Windows, Linux, macOS) and common architectures (x86, x64). LocalSolver's x64 binaries offer full 64-bit capabilities: some clients tackle models involving 50 million variables then loading 50 GB of RAM, .
Passing from LSP to APIs is easy: you have to concentrate on your math optimization model only. You do not have to decompose your problem, you do not have to tune the solver, nor even to write additional specific codes, to solve large-scale real-world optimization problems in minutes.
LocalSolver's API object model