Hybrid greedy and space-filling training#
Method. Hybrid training combines one error-indicator-selected point with additional maximin space-filling points in each batch. Validation determines when training stops.
Example. The example uses an adaptive batch to balance refinement in inaccurate regions with broader parameter-domain coverage.
Run: python examples/hybrid_greedy_demo.py --work-dir /tmp/hybrid-greedy-demo.
For algorithm details, settings, and full source listings, see hybrid greedy training — detailed notes.