Bayesian Dynamic Mode Decomposition
Bayesian Dynamic Mode Decomposition
复制标题
DOI：
10.24963/ijcai.2017/392
复制
发表时间：
2017-08
期刊：
影响因子：
--
通讯作者：
Naoya Takeishi;Y. Kawahara;Yasuo Tabei;T. Yairi
中科院分区：
文献类型：
--
作者：
Naoya Takeishi;Y. Kawahara;Yasuo Tabei;T. Yairi
动态模式分解(DMD)是一种数据驱动的计算非线性动力系统模式表示的方法，已被应用于科学和工程的各个领域。在这次演讲中，我们介绍了DMD的重构，即概率DMD和贝叶斯DMD，利用它们可以显式地合并观测噪声，对与DMD相关的量进行后验推断，并系统地考虑DMD的扩展。此外，我们还介绍了两个应用实例：贝叶斯稀疏DMD和概率DMD的混合。
Dynamic mode decomposition (DMD) is a data-driven method for calculating a modal representation of a nonlinear dynamical system, and has been utilized in various fields of science and engineering. In this talk, we introduce reformulations of DMD, namely probabilistic DMD and Bayesian DMD, with which we can explicitly incorporate observation noises, conduct posterior inference on DMD-related quantities and consider extensions of DMD in a systematic way. Furthermore, we introduce two examples of application: Bayesian sparse DMD and mixtures of probabilistic DMD.