Skip to content

Latest commit

 

History

History
76 lines (69 loc) · 2.34 KB

TODO

File metadata and controls

76 lines (69 loc) · 2.34 KB
Raw

Moo genetic algorithm library, first release [13/13]

  • [X] SBX
  • [X] UNDX
  • [X] premature convergence check (GensNoChange)
  • [X] loop-with-IO
  • [X] elite as SelectionOp
  • [X] fitness scaling as Population a -> Population a
  • [X] SelectionOp combinator
  • [X] preserve elite in crossover and mutation
  • [X] re-organize modules by coding type: binary, real-valued, permutation
  • [X] no-op mutation and crossover
  • [X] time limit
  • [X] rename Fitness to Objective; clean-up minimizing and maximizing selection ops
  • [X] don’t require to evalFitness for the initial population
  • [X] NSGA-II
  • [X] reduce boilerplate for the common default scenarios
  • [X] common framework to run GAs in all examples (CLI)
  • [X] constrained selection without penalties
  • [X] death penalty

API usability [3/3]

  • [X] withProbability to operate on arbitrary (x -> Rand x), otherwise identity
  • [X] getConstrainedGenomes(Rs) remove Rs
  • [X] getConstrainedGenomes adapt for Binary

Tests [1/4]

  • [X] Continuous problem (Rosenbrock) [4/4]
    • [X] BLX
    • [X] UNDX
    • [X] SBX
    • [X] Gaussian mutation
  • [-] Some binary problem [4/8]
    • [ ] gray encoding
    • [ ] gray real encoding
    • [ ] binary encoding
    • [X] point mutation
    • [ ] constant frequency mutation
    • [X] 1-point crossover
    • [X] 2-point crossover
    • [X] uniform crossover
  • [-] Stop conditions [4/6]
    • [X] # of generations
    • [X] GensNoChange
    • [X] fitness condition
    • [X] Or
    • [ ] And
    • [ ] time limit
  • [ ] Loop constructs [0/3]
    • [ ] loop
    • [ ] loopWithLog
    • [ ] loopIO

TODO: future releases [1/7]

  • [ ] permutation GAs, TSP example
  • [X] minimizing SelectionOp’s
  • [ ] optimize UNDX (linalg)
  • [ ] optimize SBX (linalg)
  • [ ] parallel selection
  • [ ] parallel crossover
  • [ ] parallel mutation
  • [ ] parallel objective evaluation
  • [ ] migrate to mwc-random
  • [ ] benchmark to compare operators’ performance
  • [ ] Laplace crossover (LX)
  • [ ] heuristic crossover (HX)
  • [ ] power mutation (PM)
  • [ ] Makinen, Periaux and Toivanen mutation (MPTM)

TODO: breaking changes for future releases [0/2]

  • [ ] non-list genomes
  • [ ] time-dependent mutation (generation counter)
  • [ ] Michalewicz non-uniform mutation (NUM), require time-dependend mutation