Stochastic Diffusion Search: Finding Symmetry

Too Long; Didn't Read

Stochastic Diffusion Search (SDS) is a multi-agent optimisation algorithm formed of two steps, testing and diffusion . The test phase calculates an individual’s objective value, and the diffusion phase controls decentralised communication between agents. The standard SDS algorithm comprises a population of agents with each agent’s features initialised randomly within a search space. These features serve as the agent’s hypothesis. The test phase calculates whether their hypothesis is correct or incorrect and sets each individual’s status to active or inactive based upon said hypothesis. The diffusion phase then governs communication between active and inactive agents. If an agent is inactive, it will select another agent randomly. If this selected agent is active, then the original agent will adopt it’s hypothesis as their own. However, if this chosen agent is inactive then the agent will randomise their hypothesis once more to continue exploring the search space. Many variations of the algorithm exist to adapt to different problem spaces by altering properties such as the recruitment strategy and context mechanism used.