Reproduce “Amortized Optimization for Calibrating Stochastic Social Simulators” 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
Muyao Zhong, Peng Yang
The method prepares a memory of feasible simulator settings and response statistics, then selects and optionally adjusts a candidate for each new target without another simulator call. In 75 financial windows, queries took 0.0562 ms, while a 10-minute GP-BO search had slightly lower mean error. On 64 traffic targets, local correction and tandem inverse errors were statistically indistinguishable; the inverse queried faster but required more preparation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.