create
Creates and configures a global-best particle swarm optimization (PSO) solver for a given problem definition.
Note: unlike the other create* factories (which default to sequential evaluation), this factory defaults parallelOptions to enabled = true, because PSO evaluates the whole swarm as one batch each iteration and therefore benefits directly from parallel oracle execution. As with all parallel evaluation, the supplied modelBuilder MUST return an independent, freshly built model on each call; pass ParallelEvaluationOptions() to force sequential evaluation if that is not the case.
Return
An instance of ParticleSwarmSolver that encapsulates the optimization process and results.
Parameters
The definition of the optimization problem, including constraints and objectives.
The model builder interface used to create models for evaluation. Must yield independent models per call when parallel.
The number of particles in the swarm. Defaults to ParticleSwarmSolver.defaultSwarmSize.
The inertia-weight schedule. When omitted (null), a LinearDecreasingInertia whose horizon matches maxIterations is used, so the weight decays across the whole run.
The cognitive acceleration coefficient c1. Defaults to ParticleSwarmSolver.defaultCognitiveCoefficient.
The social acceleration coefficient c2. Defaults to ParticleSwarmSolver.defaultSocialCoefficient.
How out-of-range positions are handled. Defaults to ClampToBounds.
Optional initial coordinates seeded as the first particle. If left null, the whole swarm is scattered over random feasible points upon initialization.
The maximum number of iterations the algorithm will run. Defaults to 100.
The number of replications to use during each evaluation to reduce stochastic noise.
Specifies if the evaluator uses a solution cache. By default, this is MemorySolutionCache.
Specifies if the simulation oracle will use a SimulationRunCache. The default is null (no cache).
the run parameters to apply to the model during the building process
the random number stream number; 0 (the default) means the next available stream
the provider of random number streams; defaults to a fresh RNStreamProvider, so each solver has its own streams