ParticleSystem
A Particle system is a mixture of Dirac distributions.
\[P(x) = \sum_{i= 1}^n w_i \delta(x - μ^{(i)})\]
It can also be used to represent a mixture of trajectories.
\[P(x) = \sum_{i= 1}^n w_i \delta(x_{1:T} - μ^{(i)}_{1:T})\]
Types
MarkovKernels.AbstractParticleSystem
— TypeAbstractParticleSystem{T} <: AbstractDistribution{T}
abstract type for representing systems of particles.
MarkovKernels.ParticleSystem
— TypeParticleSystem{T,A,B} <: AbstractParticleSystem{T}
type for representing a standard particle system.
Constructors
MarkovKernels.ParticleSystem
— MethodParticleSystem(logws::AbstractVector{<:Real}, X::AbstractArray{<:AbstractVector{T}})
Creates a ParticleSystem with logarithm of the mixture weights given by logws and location parameters in X. If X is an AbstractVector the resulting object represents a classical Dirac mixtrue. Whereas if X is an AbstractMatrix, the resulting object represents a Dirac mixture over trajectories, where the row dimension represents time and the column dimension enumerates the particles.
Basics
MarkovKernels.dim
— Methoddim(P::AbstractParticleSystem)
Returns the dimension of the particle system distribution P.
MarkovKernels.logweights
— Methodlogweights(P::AbstractParticleSystem)
Returns the logarithms of the mixture weights of the particle system P.
MarkovKernels.weights
— Methodweights(P::AbstractParticleSystem)
Returns the mixture weights of the particle system P.
MarkovKernels.nparticles
— Methodnparticles(P::AbstractParticleSystem)
Computes the number of particles in the particle system P.
MarkovKernels.particles
— Methodparticles(P::AbstractParticleSystem)
Returns the particle locations of the particle system P.
Statistics.mean
— Methodmean(P::AbstractParticleSystem)
Computes the mean of the particle system distribution P.