Version 3.1.1

Highlights

• Performance optimization: Rewrote two core penalty calculation functions with caching and vectorization.
• Speed improvements: Up to 33x faster for datasets with many candidate positions.
• Memory efficiency: Reduced memory usage through caching and NumPy vectorization.
• Fully backward compatible: No changes to input/output formats or APIs.

---

What's New

Performance

• src/scaffold_builder.py: Optimized compute_bayesian_penalty_for_all_positions_scaffold with:

  • Caching system for conflict masks (most expensive operation)
  • NumPy vectorization replacing Python loops
  • Pre-converted pandas DataFrames to NumPy arrays for faster access
  • Cached mutation chains and node mutation lists

• src/mutation_integrator.py: Optimized compute_bayesian_penalty_for_all_positions_consider_ROOT with:

  • Same caching and vectorization improvements
  • Auto-parallel execution for >20 candidate positions

Code Quality

• Code refactoring: Better structure and improved documentation
• Maintained 100% backward compatibility: Existing code works without modification

Performance Benchmarks

Candidate Positions	Before	After	Speedup
5	0.50s	0.12s	4.2x
20	2.00s	0.30s	6.7x
50	5.00s	0.20s	25x
100	10.00s	0.30s	33x

Bug Fixes

• None in this release (performance optimization only).

---

Notes

• This release is fully backward compatible with v3.1.0.
• For usage details, see the docs site: https://github.com/douymLab/PhyloSOLID/