Skip to content

Improve representative training batches - #73

Merged
gAldeia merged 2 commits into
masterfrom
fix/batch_size
Nov 12, 2026
Merged

gAldeia merged 2 commits into
masterfrom
fix/batch_size

Conversation

@gAldeia

@gAldeia gAldeia commented Nov 11, 2026

Copy link
Copy Markdown
Collaborator
  • Draw rotating training batches, preserving class prevalence and including every class. Repeat exhausted minority samples while completing an epoch.
  • Stratify regression batches by target quartiles.
  • Refit programs on each generation’s batch and evaluate model selection on the fixed validation partition.
  • Avoid repeated full-training evaluations and the final full-data refit.
  • Align lexicase sample weights with each batch, preserve seeded dataset splitting, and restore the DEAP evaluator binding.
  • Add C++ and Python batch tests.

This commit also ensures that batch allocation index never explode, and lexicase selection can handle batches

@gAldeia
gAldeia merged commit f57d5cc into master Nov 12, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant