- A complier compatible with c++17:
- e.g., gcc version 8.2.0 or Apple LLVM version 10.0.1 (clang-1001.0.46.4)
- IloCplex 12.8.0.0
- Boost libraries
- Python3 [inside the code, "python3" is used to run some py modules.]
- import pickle
- import sys
- import pandas as pd
- BCompose is to solve pre-decomposed problems.
- The user needs to export the master and subproblem(s) following the provided guidelines.
- This gives the user a full flexibility in decomposing the problem and exploiting its special structures.
- The exported models must be placed in 'models' directory.
- After doing this, refer to 'How To Run' to optimize the problem.
- BCompose generally works with the subproblem formulation of the generalized Benders decomposition.
- More details can be found in our paper entitled "The Benders Dual Decomposition Method".
- There are few constraint and variable naming conventions which the user needs to follow in order for BCompose to operate normally.
- For now, please refer to the provided examples under the 'examples' folder for more details on such conventions. I will soon provide a complete guideline.
- Examples on how to export the pre-decomposed problems are available for the following problems:
- Fixed-charge capacitated facility location problem:
- stochastic/deterministic version;
- complete/incomplete recourse version;
- strong/weak formulation;
- Stochastic fixed-charge multi-commodity capacitated network design problem;
- strong/weak formulation;
- Stochastic network interdiction problem;
- Covering location problem;
- Fixed-charge capacitated facility location problem:
- All the examples are created in C++.
- After exporting the proper formulations for the master (MP) and subproblem (SP) into the "models/" directory, run following command:
python3 run_me.py - or:
./BCompose --model_dir=? --current_dir=?
- Some of the main features of BCompose can be tuned via the global variables in the 'solver_setting.h' file.
#ToBeCompleted