OnePlusLambdaLambda GA

[francastagna]

PR Summary

We implemented and validated the 1+(λ, λ) Genetic Algorithm and added comprehensive tests

---

1+(λ, λ) GA Pseudocode

---

Main Changes

• New Algorithm (OnePlusLambdaLambdaGeneticAlgorithm)

  • Implements the 1+(λ, λ) loop:
    • Generate λ mutants from the parent, pick best mutant.
    • Do λ/2 crossovers between parent and best mutant, pick best offspring.
    • Replace parent iff best offspring is strictly better.
  • Lifecycle hooks: beginGeneration, endGeneration, beginStep, endStep.
  • Freezes targets per generation via frozenTargets = archive.notCoveredTargets().

• Config (EMConfig) and Wiring

  • Added OnePlusLambdaLambdaGA to EMConfig.Algorithm.
  • Uses onePlusLambdaLambdaOffspringSize (λ), xoverProbability, fixedRateMutation.

• Docs

  • Updated docs/options.md to include the new algorithm and parameters.

---

Tests

• New/Updated tests (OnePlusLambdaLambdaGeneticAlgorithmTest.kt):
  • Finds Optimum (testOnePlusLambdaLambdaGeneticAlgorithmFindsOptimum):
    • Reaches the OneMax optimum under standard constraints.
  • Parent Kept When Offspring Worse (testParentKeptWhenOffspringWorse).
  • Best Offspring Kept Better Than Parent (testBestOffspringKeptWhenBetterThanParent):
  • Edge Cases
    • xoverProbability = 0.0, fixedRateMutation = 1.0 → no crossovers; λ mutants; size preserved.
    • fixedRateMutation = 0.0, xoverProbability = 1.0 → λ/2 crossovers; 0 mutants; size preserved.

---