arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2012.09739math.NAcs.NA
收藏

使用低精度近似随机变量的舍入误差

Rounding error using low precision approximate random variables

Oliver Sheridan-Methven, Michael Giles

AI总结:

本文针对Euler-Maruyama格式的随机微分方程数值近似,提出使用低精度近似随机变量,并建模舍入误差,分析不同精度下多层蒙特卡洛的方差结构,发现单精度和半精度可分别带来约7倍和10-12倍的潜在加速。

摘要:

针对使用Euler-Maruyama格式对随机微分方程进行数值近似的问题,我们提出引入使用低精度(如单精度和半精度)计算的近似随机变量。我们提出并论证了所产生的舍入误差的模型,并针对适用于常规和嵌套多层蒙特卡洛估计的二阶和四阶差分,给出了平均情况下的误差界。通过考虑在不同精度下,有无Kahan补偿求和的多层蒙特卡洛修正项的方差结构,我们计算了各种精度可能带来的加速比。我们发现,在广泛的离散化层级上,单精度有望实现约7倍的近似加速。半精度在几个粗模拟层级上提供了相当的改进,甚至在极粗的几个层级上提供了10-12倍的改进。

英文摘要:

For numerical approximations to stochastic differential equations using the Euler-Maruyama scheme, we propose incorporating approximate random variables computed using low precisions, such as single and half precision. We propose and justify a model for the rounding error incurred, and produce an average case error bound for two and four way differences, appropriate for regular and nested multilevel Monte Carlo estimations. By considering the variance structure of multilevel Monte Carlo correction terms in various precisions with and without a Kahan compensated summation, we compute the potential speed ups offered from the various precisions. We find single precision offers the potential for approximate speed improvements by a factor of 7 across a wide span of discretisation levels. Half precision offers comparable improvements for several levels of coarse simulations, and even offers improvements by a factor of 10-12 for the very coarsest few levels.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑