Inspired from quantum Monte Carlo methods, we developed a novel, fast, accurate, robust, and generalizable high performance algorithm for Monte Carlo Parametric Expectation Maximization (MCPEM) methods. We named it Randomized Parametric Expectation Maximization (RPEM). RPEM can be used on a personal computer as an independent engine or can serve as a `booster' to be combined with MCPEM engines used in current population modeling software tools. RPEM can also run on supercomputer clusters, since it is fully parallelized and scalable.
翻译:在量子蒙特卡洛方法的启发下,我们为蒙特卡洛预期最大化(MCPEM)方法开发了新型的、快速的、准确的、稳健的和可通用的高性能算法,我们将其命名为随机的参数预期最大化(RPEM)。 RPEM可以作为独立引擎在个人计算机上使用,也可以作为“加速器”与MCPEM当前人口模型软件工具中使用的发动机结合使用。 RPEM也可以在超级计算机集群上运行,因为它是完全平行和可扩缩的。