Parallel Particle Learning for Bayesian Asset Price Prediction

• Activity Number: 352
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Kenichiro McAlinn*+ and Teruo Nakatsuma and Hiroaki Katsura
• Companies: Keio University and Keio University and Keio University
• Keywords: Bayesian Inference ; Financial Analysis ; Parallel Computing ; Particle Learning

Abstract

Posterior simulation for Bayesian inference using particle filters and particle learning algorithms have proven to be successful in potently predicting asset prices and other financial measures like volatility. However, these particle based methods for posterior simulation are, by nature, computationally strenuous and more so as the model becomes more complicated. This fact has stymied its success in the practical world and is yet to replace MCMC algorithms for many researchers. However, with the recent development of fast and inexpensive devices for parallel computing, such as general purpose graphic processing units (GPGPU), formerly impractical computations, that would take hours or even days, can be completed in minutes or even seconds. With this new paradigm in mind, we have developed a new algorithm for particle learning that is fully parallelized and thus making it suitable for GP