COIN-OR

COIN-OR
Name	COIN-OR
Developer	INFORMS community, academics, industry contributors
Released	2000s
Programming language	C++, Python, Fortran
Operating system	Linux, Microsoft Windows, macOS
Genre	Optimization software
License	Eclipse Public License

COIN-OR is an open-source initiative providing a repository of software for computational optimization applications. It offers a collection of libraries, solvers, and modeling tools used by researchers, practitioners, and educators across operations research, finance, energy, and transportation sectors. The project emphasizes reusable components, academic collaboration, and deployment in industrial settings.

Overview

COIN-OR aggregates a suite of projects focused on mathematical optimization techniques such as linear programming, integer programming, nonlinear programming, and stochastic programming. The ecosystem includes solver engines, preprocessing utilities, modeling interfaces, and algorithmic building blocks that integrate with platforms like AMPL, GAMS, Julia, Python, and Microsoft Excel. It aims to bridge academic research from institutions such as the Massachusetts Institute of Technology, University of California, Berkeley, Stanford University, and New York University with industry adopters including Amazon, General Electric, Siemens, and IBM.

History and Development

The initiative emerged during the early 2000s as academics and practitioners sought open alternatives to proprietary systems like CPLEX and Gurobi. Early contributors included researchers affiliated with Carnegie Mellon University, University of Cambridge, and Northwestern University who ported algorithms from conferences such as the INFORMS Annual Meeting, SIAM Optimization Conference, and journals like Mathematical Programming. Over time, funding, contributions, and governance have involved partnerships with societies such as INFORMS and collaborations with government labs such as Los Alamos National Laboratory and Sandia National Laboratories.

Software Projects and Components

Key components in the repository include solver libraries and modeling tools. Notable projects are optimization solvers comparable to GLPK and components similar in scope to CoinMP-style interfaces: linear solvers, mixed-integer solvers influenced by algorithms from Egon Balas-type branch-and-bound, and presolve techniques discussed in venues like the International Symposium on Mathematical Programming. Modeling and API layers tie into languages and tools such as AMPL, GAMS, R, Python, and Julia. The collection supports algorithmic components inspired by work published in Operations Research and Management Science and leverages numerical libraries akin to BLAS, LAPACK, and solver frameworks used by NASA mission planning and Google-scale scheduling research.

Architecture and Technical Features

The architecture emphasizes modular C++ libraries with bindings to high-level languages like Python and Java. Core technical features include sparse matrix representations influenced by standards used in NETLIB benchmarks, iterative linear solvers comparable to routines in ARPACK, and decomposition techniques analogous to Benders decomposition and Dantzig–Wolfe decomposition. Interaction patterns resemble those in projects hosted on platforms such as GitHub and continuous integration practices from Travis CI and Jenkins. The design supports parallelism approaches related to research from Lawrence Livermore National Laboratory on distributed optimization and interfaces for solver comparison common at the MIPLIB benchmarks.

Licensing and Governance

The initiative employs permissive open-source licensing models similar to the Eclipse Public License and engages governance practices resembling working groups in INFORMS and SIAM. Contributors hail from academic institutions like University of Oxford and corporations such as Microsoft and Amazon, adhering to contribution guidelines that reflect norms used by projects under the Apache Software Foundation and Linux Foundation umbrellas. Decision-making processes have parallels to editorial boards for journals like Mathematical Programming and steering committees found in organizations such as ACM.

Users and Applications

End users include researchers at Princeton University, practitioners at Deloitte, engineers at Siemens, and data scientists at Google and Facebook. Application domains cover supply chain planning for companies like Walmart, power systems optimization studied by National Renewable Energy Laboratory, airline crew scheduling investigated at Delta Air Lines, and portfolio optimization in financial firms such as Goldman Sachs. The software has been cited in case studies presented at venues like the INFORMS Annual Meeting and adopted in coursework at universities including Harvard University and University of Michigan.

Community and Contribution Model

The community model mirrors collaborative projects from GitHub and academic consortia, combining contributions from faculty, graduate students, and industry engineers. Communication occurs through mailing lists, issue trackers and code review workflows similar to those used by OpenAI and large open-source projects. Contribution paths include submitting patches, documentation comparable to materials in SIAM Review, and participating in workshops at conferences like the Workshop on Algorithm Engineering and Experiments and the International Conference on Integer Programming and Combinatorial Optimization.

Category:Optimization software Category:Open-source software