This paper considers Event-Chain Monte Carlo simulation schemes in order to design an original irreversible Markov Chain Monte Carlo (MCMC) algorithm for the sampling of complex statistical models. The functioning principles of MCMC sampling methods are firstly recalled, as well as standard Event-Chain Monte Carlo simulation schemes are described. Then, a Forward Event-Chain Monte Carlo sampling methodology is proposed and introduced. This nonreversible MCMC rejection-free simulation algorithm is tested and run for the sampling of high-dimensional ill-conditioned Gaussian statistical distributions. Numerical experiments demonstrate the efficiency of the proposed approach, compared to standard Event-Chain and standard Monte Carlo sampling methods. Accelerations up to several magnitudes are exhibited.
View on arXiv