GDnD Wiki

GDnD Wiki

This site is for Ulster University students and staff. Enter the class access password to continue.

Natural Computing Algorithms

2 min read

Source metadata

  • Type: Survey textbook
  • Authors: Anthony Brabazon, Michael O’Neill, Seán McGarraghy
  • Published: 2015, Springer

Key takeaways

  • Establishes a common conceptual frame across natural computing: representation, fitness/utility, variation, and selection/update dynamics.
  • Gives a clean formulation of the canonical genetic algorithm, including genotype/phenotype distinction, population update loop, and operator choices.
  • Explains particle swarm optimisation as search driven by both personal memory (pbest) and social memory (gbest), making it a strong complement to GA rather than a replacement.
  • Surveys ant-inspired algorithms as pheromone-based constructive search methods, especially relevant to routing and path construction problems.
  • Presents neuroevolution as the use of evolutionary search to discover neural network inputs, structures, and/or weights when gradient-based tuning is awkward or unavailable.

Notable claims

  • Evolutionary algorithms act on encodings of solutions, not directly on solutions themselves; the phenotype must be decoded before fitness is measured.
  • PSO’s key advantage over random search is that it preserves both individual memory and social communication when choosing the next move.
  • Neuroevolution is valuable because neural-network design choices create huge combinatorial search spaces that are difficult to tune manually.

Relevance

Directly informs:

Supports:

Open questions raised

  • Which of these methods deserve full Unity/C# worked examples in code/ rather than wiki-only treatment?
  • For teaching purposes, when is PSO easier to communicate than GA, and when is it not?

genetic-algorithms · particle-swarm-optimisation · ant-colony-optimisation · neuroevolution · machine-learning-games · overview-artificial-intelligence-in-games

Graph View

Backlinks