|
vSMC
vSMC: Scalable Monte Carlo
|
SingleParticle now as iterator-like behaviors.Particle now has a new sp member that return SingleParticle objects.NormalMVDistribution and related functions for generating multivariate Normal distribution random variate.Covariance is a new class that can compute the covariance matrix from (weighted) samples, and output the matrix in variance formats.cov_chol transform a covariance matrix, stored in various formats, into the lower triangular part of its Cholesky decomposition, stored as a packed row major matrixAlignedAllocator template parameter Memory now requires its member functions to be static and the class stateless. Unless you write your allocator using this template, there is no need to change client code.MatrixOrder is now an alias to MatrixLayout.U01LRDistribution is removed, while U01CCDistribution etc., remain.UniformRealLRDistribution and UniformRealCCDistribution etc., are removed.Monitor. See the GMM example.http://zhouyan.github.io/vSMCDoc/v2.2.0/
The mutation (MCMC) steps are now performed after the resampling of initialization step.
http://zhouyan.github.io/vSMCDoc/v2.1.0/
The version 2.0.0 is a major restructuring of the library since the initial release. Its documentation is still been updated. For the time being, one can keep using the old JSS release or the reference manual below.
The most important change is that C++11, both language and libraries are now required.
1.8.11