Click or drag to resize

Altaxo.Calc.Statistics.Mcmc Namespace

[Missing <summary> documentation for "N:Altaxo.Calc.Statistics.Mcmc"]

Classes
 ClassDescription
Public classHybridMC A hybrid Monte Carlo sampler for multivariate distributions.
Public classHybridMCGenericT The Hybrid (also called Hamiltonian) Monte Carlo produces samples from distribution P using a set of Hamiltonian equations to guide the sampling process. It uses the negative of the log density as a potential energy, and a randomly generated momentum to set up a Hamiltonian system, which is then used to sample the distribution. This can result in a faster convergence than the random walk Metropolis sampler ().
Public classMCMCDiagnostics Provides utilities to analysis the convergence of a set of samples from a .
Public classMcmcSamplerT The interface which every sampler must implement.
Public classMetropolisHastingsSamplerT Metropolis-Hastings sampling produces samples from distribution P by sampling from a proposal distribution Q and accepting/rejecting based on the density of P. Metropolis-Hastings sampling doesn't require that the proposal distribution Q is symmetric in comparison to . It does need to be able to evaluate the proposal sampler's log density though. All densities are required to be in log space. The Metropolis-Hastings sampler is a stateful sampler. It keeps track of where it currently is in the domain of the distribution P.
Public classMetropolisSamplerT Metropolis sampling produces samples from distribution P by sampling from a proposal distribution Q and accepting/rejecting based on the density of P. Metropolis sampling requires that the proposal distribution Q is symmetric. All densities are required to be in log space. The Metropolis sampler is a stateful sampler. It keeps track of where it currently is in the domain of the distribution P.
Public classRejectionSamplerT Rejection sampling produces samples from distribution P by sampling from a proposal distribution Q and accepting/rejecting based on the density of P and Q. The density of P and Q don't need to to be normalized, but we do need that for each x, P(x) < Q(x).
Public classUnivariateHybridMC A hybrid Monte Carlo sampler for univariate distributions.
Public classUnivariateSliceSampler Slice sampling produces samples from distribution P by uniformly sampling from under the pdf of P using a technique described in "Slice Sampling", R. Neal, 2003. All densities are required to be in log space. The slice sampler is a stateful sampler. It keeps track of where it currently is in the domain of the distribution P.
Delegates
 DelegateDescription
Public delegateDensityT A function which evaluates a density.
Public delegateDensityLnT A function which evaluates a log density.
Public delegateGlobalProposalSamplerT A method which samples datapoints from a proposal distribution. The implementation of this sampler is stateless: no variables are saved between two calls to Sample. This proposal is different from in that it doesn't take any parameters; it samples random variables from the whole domain.
Public delegateHybridMCGenericTDiffMethod The delegate type that defines a derivative evaluated at a certain point.
Public delegateLocalProposalSamplerT A method which samples datapoints from a proposal distribution given an initial sample. The implementation of this sampler is stateless: no variables are saved between two calls to Sample. This proposal is different from in that it samples locally around an initial point. In other words, it makes a small local move rather than producing a global sample from the proposal.
Public delegateTransitionKernelLnT A function which evaluates the log of a transition kernel probability.