NEWS.md

In AlexisDerumigny/MMDCopula: Robust Estimation of Copulas by Maximum Mean Discrepancy

MMDCopula 0.2.1

NEW FEATURES

• New functions BiCopPar2Tau.MO and BiCopTau2Par.MO to convert between the parameter and the Kendall's tau of a Marshall-Olkin copula.

• New argument truncVal of the function BiCopParamDistLp to allow for computation of the norm on a smaller subset of the unit square [0,1]^2. BiCopParamDistLp now also works for p=Inf (i.e., the supremum norm).

• New checks for finiteness of the arguments to BiCopEstMMD.

OTHER IMPROVEMENTS

• Improvement of default values for gamma in BiCopEstMMD.

• Updated the references.

• Fixed the documentation of the parameter kernel for the functions BiCopEstMMD and BiCopGradMMD.

• New dependence wdm instead of pcaPP for fast computation of Kendall's tau.

MMDCopula 0.2.0

• Change email address of maintainer Alexis Derumigny corresponding to new affiliation.
• Vectorized version of BiCopSim.MO.
• BiCopSim.MO now returns a list with the estimated parameter and Kendall's tau for consistency with the other BiCopEstMMD function.
• New function BiCopMMDConfInt for bootstrapped- and subsampling-based confidence intervals.
• Renaming of kernels such as gaussian.KG to gaussian.Phi for coherence.
• Improvement of the stochastic gradient interface.
• Better handling of limiting cases (when Kendall's tau is close to the boundary).

MMDCopula 0.1.0

• Initial release

AlexisDerumigny/MMDCopula documentation built on April 27, 2022, 2:20 a.m.