Evergreen Ballroom Prefunction
Detecting Pair-Copula Dependence Using Convolutional Neural Network (306689)*Xiao Yuan, Purdue University Fort Wayne
Keywords: Pair-wise copula, dependence, convolutional neural network
Modelling dependence in high-dimensional data is often a challenging task due to the multivariate nature and complex dependence structure. Recently, vine copula has become popular in modeling such data: it uses pair-wise copula as basic "building block" to construct a hierarchical structure to model the likelihood function. Since the pair-copula is the fundamental component in the building of the hierarchy, accurate detection of the pair-wise dependence becomes crucial. In this project, convolutional neural network is used to recognize the dependence pattern in pair copula for better accuracy, and neural network architecture is also investigated for optimal performance.