Active learning

[breimanntools]

Problem

Mutation selection is static and does not adapt based on model uncertainty or feature space coverage.

Goal

Enable active learning to select the most informative mutation candidates for iterative experiments.

Tasks

• Implement uncertainty-based selection (e.g. model confidence)
• Implement diversity-based selection (feature space coverage)
• Avoid redundant mutation candidates
• Rank candidates by informativeness
• Integrate with mutation suggestion workflow

How this improves AAanalysis

• Reduces number of required experiments
• Improves efficiency of iterative design
• Enables smarter exploration of sequence space

Acceptance criteria

• Reduces candidate set size while maintaining diversity
• Improves learning efficiency across iterations
• Fully integrates with CPP feature space