Tutorials

Staged walkthroughs, from a minimal continuous-optimization run to a full config-driven, multi-objective setup, ending with how to build your own assembler/fitness function/launcher from scratch. Every code snippet has been run against a real navicatGA install and its output verified, except where a page says otherwise (Tutorial 6 needs AaronTools, not installed while writing these). Copy-paste them as a starting point.

• 1. Continuous optimization
• 2. SMILES fragment optimization
• 3. Editing a real molecule (SELFIES)
• 4. Config-driven runs
• 5. Multi-objective optimization
• 6. 3D structures (XYZ/AaronTools)
• 7. Writing your own assembler, fitness function, and launcher
• 8. Logging and saving run artifacts

Tutorial	You’ll learn
1. Continuous optimization	The core assembler/fitness_function pattern with FloatGenAlgSolver; GenAlgSolver’s shared GA-mechanics params.
2. SMILES fragment optimization	Alphabets, excluded_genes, seeding a starting population, SmilesGenAlgSolver.
3. Editing a real molecule (SELFIES)	Why SELFIES over SMILES; freezing a scaffold and mutating only new positions.
4. Config-driven runs	Moving a Python constructor call into a YAML file with navicatGA.config, using templates/.
5. Multi-objective optimization	Returning several fitness values and combining them with a scalarizer (chimera).
6. 3D structures (XYZ/AaronTools)	Where XYZGenAlgSolver fits, and its extra dependency.
7. Writing your own assembler/fitness/launcher	Building assemble()/score() from scratch for a custom problem, and the launcher patterns for runtime-only values and multi-cycle runs.
8. Logging and saving run artifacts	The standard procedure for deciding what a run writes and who writes it: run directories, fitness-side intermediates, per-cycle trajectories, and what the solver does not keep.