Arne Schulze (Chair of Business administration, HSU)
A Mixed-integer programming (MIP) model solved by the Branch-a-Bound (B&B) method is the main exact solution strategy in combinatorial optimization like in production and logistics. The aforementioned optimization problems are typically NP-hard to solve such that a vast amount of computation power is required to solve instances of reasonable size to optimality. There are basically three options to improve the search: branching, bounding, and parallelization. While bounding is strongly problem-specific, the other two can be addressed more generally which is the aim of the project. Efficient branching strategies and parallelization are elementary for a successful application of the B&B method. B&B is a standard approach implemented in state-of-the-art solvers like Gurobi or CPLEX. These solvers allow for user interaction by adding cuts, developing branching strategies or parallelize the solution procedure. In the project, a self-learning branching strategy shall be evaluated and be benchmarked in an HPC environment against the standard best-bound branching strategy. Thereby, a main focus is given on parallelization to accelerate the search process.