Examples

The Evolver project includes many runnable examples, in the org.uma.evolver.example package:

• org.uma.evolver.example.baselevel: configurable algorithms used on their own, without meta-optimization.

  • standard: typical configurations of every algorithm (e.g. NSGAIIForZDT1Example, MOEADBiObjectiveTSPExample).

  • tuned: configurations found by meta-optimization (e.g. NSGAIIBiObjectiveTSPExample).

  • features: demonstrations of specific capabilities, such as external archives or observers (e.g. NSGAIIZDT4WithArchiveExample).

• org.uma.evolver.example.training: meta-optimization runs, grouped by benchmark (zdt, dtlz, re3d, tsp). They cover the flat and tree encodings and several meta-optimizers (NSGA-II, SPEA2, SMPSO, Async NSGA-II, Random Search, …); most of them build a TrainingRequest and run it with TrainingRunner, the same pipeline used by cli.training.

• org.uma.evolver.example.validation: comparative studies of tuned and standard configurations on validation problems.

Most training examples can also be run without Java code from the request.yaml files in src/main/resources/cli/training/ (see CLI Tools).