A General Purpose Approximation to the Ferguson-Klass Algorithm for Sampling from Lévy Processes Without Gaussian Components
提出一种通用方法，通过网格上跳跃强度的多部分近似并应用Ferguson-Klass算法，从无高斯成分的Lévy过程中高效抽样，速度比原算法快几个数量级，适用于超越共轭假设的贝叶斯非参数模型。
We propose a general-purpose method for generating samples from Lévy processes without Gaussian components. It uses a multi-part approximations of the jump intensity on a grid and applies the Ferguson-Klass algorithm. We consider how the choice of grid affects the approximation error and propose adaptive selection methods that lead to negligible approximation error. The proposed method is shown to be orders of magnitude faster than the original Ferguson-Klass algorithm and competitive with tailored methods. The method opens an avenue for computationally efficient and scalable Bayesian nonparametric models which go beyond conjugacy assumptions, as demonstrated in the examples section.