Batch space-filling ROM training#

Method. Space-filling training selects parameter points to cover the domain without relying on a ROM error indicator. New full-order simulations are run in batches and the ROM is validated after each batch.

Example. The example trains a Gaussian-process surrogate on an irregular disk-shaped parameter domain using file-backed model evaluations and cross-validation.

Run: python examples/space_filling_demo.py --work-dir /tmp/space-filling-demo.

For algorithm details, settings, and full source listings, see space filling training — detailed notes.