Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 387 - Software
Type: Contributed
Date/Time: Thursday, August 12, 2021 : 12:00 PM to 1:50 PM
Sponsor: Section on Statistical Computing
Abstract #318878
Title: Approximation of Infinite Sums by Adaptive Truncation: Applications to Statistics
Author(s): Luiz Max Carvalho* and Guido Alberti Moreira
Companies: School of Applied Mathematics, Getulio Vargas Foundation and Department of Statistics, Federal University of Minas Gerais
Keywords: marginalisation; infinite series; discrete latent variables; Hamiltonian Monte Carlo; truncation
Abstract:

Marginalising over discrete latent random variables has become more relevant with the popularization of Hamiltonian Monte Carlo techniques in the Bayesian inference context, which only allow for continuous variables. For many major infinite series, custom algorithms have been developed which exploit specific features of each problem. General techniques, suitable for a large class of problems with limited input from the user are somewhat scant, however. Here we employ basic results from the theory of infinite series to investigate general, problem-agnostic algorithms to approximate these sums to within an arbitrary tolerance $\epsilon > 0$. We compare three tentative solutions to estimating the infinite sum of interest and obtain simple theoretical results which show under which conditions each technique succeeds in providing a truncated sum within the required tolerance. We provide a host of statistical applications where truncation is necessary and compare the error achieved by each approach, as well as the number of function evaluations necessary for each one. A comparison based on computation time is also provided.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2021 program