Faster estimation of High-Dimensional Vine Copulas with Automatic Differentiation
نویسندگان
چکیده
Vine copula is an important tool in modeling dependence structures of continuousvalued random variables. The maximum likelihood estimation (MLE) for vine copulas has long been considered computationally difficult in higher dimensions, even in 10 or 20 dimensions. Current computational practice, including the implementation in the state-ofthe-art R package VineCopula, suffers from the bottleneck of slow calculations of gradients for the log-likelihood function. We show that by using techniques and tools developed in automatic differentiation, gradients of the log-likelihood function can be accurately calculated in orders of magnitude faster than the current practice. This change immediately accelerates the MLE for vine copulas by 1-2 orders of magnitudes in moderately high dimension. Furthermore, implementation with automatic differentiation is much simpler, e.g., only the objective function evaluation needs to be implemented, and higher order derivatives are readily available. We implement the case of D-Vines with Clayton copulas, and demonstrate such advantages over the current implementation in VineCopula. The purpose of this paper is to further advocate the usage of automatic differentiation tools, which appear to be less known or used in the statistical community.
منابع مشابه
Nonparametric estimation of simpli ed vine copula models: comparison of methods
Thomas Nagler*, Christian Schellhase, and Claudia Czado Nonparametric estimation of simpli ed vine copula models: comparison of methods DOI 10.1515/demo-2017-0007 Received December 27, 2016; accepted May 16, 2017 Abstract: In the last decade, simpli ed vine copula models have been an active area of research. They build a high dimensional probability density from the product of marginals densiti...
متن کاملGaussian Process Vine Copulas for Multivariate Dependence
Copulas allow to learn marginal distributions separately from the multivariate dependence structure (copula) that links them together into a density function. Vine factorizations ease the learning of high-dimensional copulas by constructing a hierarchy of conditional bivariate copulas. However, to simplify inference, it is common to assume that each of these conditional bivariate copulas is ind...
متن کاملTail dependence functions and vine copulas
Tail dependence and conditional tail dependence functions describe, respectively, the tail probabilities and conditional tail probabilities of a copula at various relative scales. The properties as well as the interplay of these two functions are established based upon their homogeneous structures. The extremal dependence of a copula, as described by its extreme value copulas, is shown to be co...
متن کاملModel selection for discrete regular vine copulas
Abstract Discrete vine copulas, introduced by Panagiotelis et al. (2012), provide a flexible modeling framework for high-dimensional data and have significant computational advantages over competing methods. A vine-based multivariate probability mass function is constructed from bivariate copula building blocks and univariate marginal distributions. However, even for a moderate number of variab...
متن کامل