Login
Login

SuanShu 3.0.0 is now released

Home 21090308 Forums General Category General Discussion SuanShu 3.0.0 is now released

  • This topic is empty.
Viewing 1 post (of 1 total)
  • Author
    Posts
  • #2639
    Ryu
    Member

    We have released an update of SuanShu, the Java numerical and statistical library. http://www.numericalmethod.com/trac/numericalmethod/wiki/SuanShu
    This is a major release.

    Other highlights are as the following.

    Basic

    • improved the performance of various core linear algebra operations, e.g., eigen decomposition, singular value decomposition, QR decomposition, Cholesky decomposition, Householder reflection, forward/backward substitution, matrix inverse, etc.
    • improved the performance of sparse matrix package
    • added new features in stats/random package: ConcurrentStandardNormalRNG, HypersphereRVG, ThinRNG/RVG, BurnInRNG/RVG, random vector generation, etc.
    • added new features in stats/distribution package: TriangularDistribution, MultinomialDistribution, DirichletDistribution, etc.
    • added DQDS and MR3 algorithms for efficient eigen decomposition of symmetric matrix
    • fixed issues in PositiveDefiniteMatrixByPositiveDiagonal when the range of eigenvalues is too large
    • fixed stack overflow issue in LPBoundedMinimizer
    • other minor bugfixes

    Optimization

    • added CSDPMinimizer for solving SDP problems
    • added QPDualActiveSetMinimizer for solving QP problems
    • fixed multiple bugs in HomogeneousPathFollowingMinimizer
    • fixed divide-by-zero issue in QPSimpleMinimizer
    • fixed stack overflow issue in QPPrimalActiveSetMinimizer

    Stats

    • added ‘mcmc’ package for sampling from distributions
    • added new features in stats/random/variancereduction: AntitheticVariates, ControlVariates, CommonRandomNumbers, ImportanceSampling, etc.
    • added UnconstrainedLASSObyCoordinateDescent for solving LASSO problems
    • added CovarianceSelectionGLASSOFAST for covariance selection
    • fixed bugs in CointegrationMLE
Viewing 1 post (of 1 total)
  • You must be logged in to reply to this topic.