Mining

A Zhen miner is a calibration engine: receive training measurements, search a ~5-parameter space for RC values that fit them, and submit parameters back before the round deadline.

Summary

  • Optimizer is your choice — Bayesian optimization, CMA-ES, MCMC, surrogate-assisted, or hybrid. No single method dominates every building type.
  • Simulation budget — up to 1,000 RC-model evaluations per round. Convergence weight rewards using fewer.
  • Local score replay — the open-source scoring module lets you validate exactly what the validator will compute before you submit.
  • Default-parameter detection zeroes out lazy submissions; there is no value in betting on priors.
  • Hardware — miner compute is CPU-bound. Budget a modern multi-core machine; no GPU required.

Canonical source