Bayesian optimisation for fast and safe parameter tuning of SwissFEL
Bayesian optimisation for fast and safe parameter tuning of SwissFEL
复制标题
用于快速、安全调整 SwissFEL 参数的贝叶斯优化
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
A. Adelmann
中科院分区：
文献类型：
--
作者：
Johannes Kirschner;Manuel Nonnenmacher;Mojmír Mutný;Andreas Krause;N. Hiller;R. Ischebeck;A. Adelmann
在加速器设施中，参数调整是一项非常耗时的任务。作为噪声评估的全局优化工具，贝叶斯优化最近被证明优于其他方法。通过使用所有可用数据学习底层函数的模型，可以仔细选择下一个评估，以尽可能少的步骤和不违反任何安全约束来确定最佳值。然而，每步计算时间随着参数的数量而显著增加，并且该方法的通用性可能导致更容易优化的函数收敛缓慢。为了克服这些局限性，我们将全局问题划分为连续的子问题，这些子问题可以使用安全的贝叶斯优化来快速解决。这使我们能够在局部和全局收敛的情况下进行交易，并适应目标函数中的附加结构。此外，我们还提供了函数的切片图作为优化过程中的用户反馈。我们展示了如何使用我们的算法来同时调整多达40个参数的SwissFEL的FEL输出，并在30分钟（< 2000步）的合理调整时间内达到收敛。
Parameter tuning is a notoriously time-consuming task in accelerator facilities. As tool for global optimization with noisy evaluations, Bayesian optimization was recently shown to outperform alternative methods. By learning a model of the underlying function using all available data, the next evaluation can be chosen carefully to ind the optimum with as few steps as possible and without violating any safety constraints. However, the per-step computation time increases signiicantly with the number of parameters and the generality of the approach can lead to slow convergence on functions that are easier to optimize. To overcome these limitations, we divide the global problem into sequential subproblems that can be solved eiciently using safe Bayesian optimization. This allows us to trade of local and global convergence and to adapt to additional structure in the objective function. Further, we provide slice-plots of the function as user feedback during the optimization. We showcase how we use our algorithm to tune up the FEL output of SwissFEL with up to 40 parameters simultaneously, and reach convergence within reasonable tuning times in the order of 30 minutes (< 2000 steps).