Reproduce “Learned Preconditioning for a Primal-Dual Interior-Point Method” 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
Abhinav Madabhushi, Jialin Liu, Minxin Zhang
pdLIP learns a positive coordinate-wise scaling for the primal update in a primal–dual optimizer, avoiding Newton-system solves during learned iterations while retaining analytical updates for the other directions. It then warm-starts pdProj for refinement. Across four 200-variable benchmark classes, at a 10⁻⁸ KKT-residual threshold, the pipeline reduced refinement iterations by 63.38%–67.10% versus cold-started pdProj; this is not a guarantee of high-accuracy solutions from pdLIP alone.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.