Reproduce “Conformal Prediction at Performative Equilibrium: The Price of Performativity” 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
Ahmed Mahrous, Roberto Di Pietro
P-PCP retrains on data induced by its current predictor, then calibrates prediction sets on fresh held-out data after fixing the final predictor. This transfers conformal coverage near equilibrium, with additional loss bounded by L times distance to equilibrium. For L≥1, no finite-sample method uniformly guarantees coverage and nontrivial sets; below 1, informative excess miscoverage has a lower bound proportional to 1/((1−L)√n).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.