Reproduce “EnerGNN: Learning Optimization-Compatible Energy Functions for Exact Constrained Combinatorial Inference” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Niki Triantafyllou, Maria M. Papathanasiou
EnerGNN learns a score for complete decision configurations and uses a mixed-integer optimization solver to choose the lowest-scoring option satisfying encoded constraints. In CART, it reports a 0.05% mean primal gap, 100% feasibility, and mean online runtime of 214.03 seconds, versus 3,728.82 seconds for full MIP. Exact optimization applies to learned energy, not the original objective, whose ranking it only approximates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.