Exploiting task-based parallelism in Bayesian Uncertainty Quantification

Abstract

We introduce a task-parallel framework for non-intrusive Bayesian Uncertainty Quantification and Propagation of complex and computationally demanding physical models on massively parallel computing architectures. The framework incorporates Laplace asymptotic approximations and stochastic algorithms along with distributed numerical differentiation. Sampling is based on the Transitional Markov Chain Monte Carlo algorithm and its variants while the optimization tasks associated with the asymptotic approximations are treated via the Covariance Matrix Adaptation Evolution Strategy. Exploitation of task-based parallelism is based on a platform-agnostic adaptive load balancing library that orchestrates scheduling of multiple physical model evaluations on computing platforms that range from multicore systems to hybrid GPU clusters. Experimental results using representative applications demonstrate the flexibility and excellent scalability of the proposed framework.