PlasmoAlgorithms.jl

PlasmoAlgorithms is a library of decomposition approaches in Julia that have been implemented for Plasmo.jl data structures. Plasmo is an open-source modeling package that constructs graph-structured optimization models. Plasmo is built on an abstraction called an OptiGraph, containing OptiNodes (with objectives, variables, and constraints) and OptiEdges (with constraints for variables on two or more nodes).

PlasmoAlgorithms is intended to include different decomposition approaches that can exploit the graph structure of an optimization, and the implementation of these algorithms is based upon the graph structure. Currently, the following are implemented in this repository:

• Benders Decomposition is implemented through the package PlasmoBenders.jl that is contained in this repository. PlasmoBenders.jl applies Benders and Nested Benders Decompositions to a problem defined by the user in Plasmo. PlasmoBenders applies these decomposition approaches based on the structure defined by the graph, where the graph structure dictates the complicating variables for the algorithm. Benders can be used for LP and MILP problems.
• Overlapping Schwarz Decomposition is implemented throught he package PlasmoSchwarz.jl (formerly SchwarzOpt.jl). PlasmoSchwarz.jl uses a partitioned graph and overlaps the partitioned regions with their neighboring partitions. Each overlapped subgraph is optimized and information is then shared between the subgraphs. Overlapping Schwarz can be used for NLP problems.

Other decomposition approaches could likewise be developed. We welcome contributions to the PlasmoAlgorithms suite. In addition, we encourage reporting issues and feature requests via the GitHub issue tracker..