DecompBench Method Selector

Feature-calibrated method selector for large-scale optimization decomposition.

Intended Use

Recommends a decomposition strategy (Benders, Dantzig-Wolfe, Column Generation, Lagrangian, Progressive Hedging, ADMM) from instance structure features.

Input Features

• n_master_variables, n_sub_variables, n_coupling_constraints
• n_scenarios, block_angularity, constraint_density, decomposability_score

Artifacts

• selector_weights.json — scoring weights
• config.json — model metadata

Limitations

Calibrated on synthetic instances. Validate on production models before deployment.

Space

https://huggingface.co/spaces/alirezaaminzadeh/large-scale-optimization-decomposition-lab