Linear-Genetic-Programming

Implementation of Linear Genetic Programming.

Introduction

This project is the result of study and try doing. While it works, it is not ready for production yet.

Work Done

• Learner

  • Evolution via Tournament, 4 indevidous go in and maintains the 2 best. On the other 2 tries to do Crossover or Mutation until find indeviduos with different Effective Instructions Sets
  • Evolution via Island. This works over other learners. The idea is to divide the full populations in smaller populations. The islands are organized in a ring and slowly let some individuos pass to the next island on the ring. This allow some local searchs in diferent areas of the search space.

• Mutations

  • MutationAddCommand: adds a random instruction
  • MutationDeleteCommand: removes a random instruction
  • MutationRandomPoint: changes an instruction with a random instruction
  • MutationMicro: slitle changes to change an existing instruction. For now just the register index is changed, in the future the operation shall also be mutable

• Crossover (if we really can call it like that)

  • CrossoverRandom: a subset of instructions are changed with another individuo (usually with a very small size)
  • CrossoverEffectiveRandom: a subset of effective instruction are changed with another individuo (usualy with a very small size)
  • CrossoverHomogeneous: a subset of instructions are changed with an equal location/size sequence of instruction of another individuo

  Work to Do

  • MutationMicro shall be able to change the instruction leaving the register unchanged
  • Do Unit Tests
  • Track changes on each indeviduo