Brief survey on computational solutions for Bayesian inference
JD Alves, João Filipe Ferreira, Jorge Lobo, Jorge M.M. Dias · Nottingham Trent University's Institutional Repository (Nottingham Trent Repository) · 2015
In this paper, we present a brief review of research work attempting to tackle the issue of tractability in Bayesian inference, including an analysis of the applicability and trade-offs of each proposed solution. In recent years, the Bayesian approach has become increasingly popular, endowing autonomous systems with the ability to deal with uncertainty and incompleteness. However, these systems are also expected to be efficient, while Bayesian inference in general is known to be an NP-hard problem, making it paramount to develop approaches dealing with this complexity in order to allow the implementation of usable Bayesian solutions. Novel computational paradigms and also major developments in massively parallel computation technologies, such as multi-core processors, GPUs and FPGAs, provide us with an inkling of the roadmap in Bayesian computation for upcoming years.