Active learning in nanoparticles: Expected Improvement picks 10 experiments per iteration

Set the GP nugget to measured replicate variance from 3 replicates, not 1e-6, and let Expected Improvement choose 10 nanoparticle experiments per iteration. Independent technical note mirroring the canonical version: https://nano-matter.com/blog/active-learning-in-nanoparticles-expected-improvement-picks-10-experiments-per-iteration.php