Second Order Adjoints

Second Order Adjoints
复制标题

二阶伴随词

DOI:
--
复制
发表时间:
--
期刊:
影响因子:
--
通讯作者:
Adrian Sandu
Adrian Sandu
中科院分区:
--
文献类型:
--
作者:
Adrian Sandu

文献摘要

被引文献

相似文献

一阶伴随给出了代价泛函关于状态的(一)阶导数。二阶伴随给出了代价泛函关于状态的二阶导数。这些导数对于加速资料同化中的优化过程和计算Hessian奇异向量是有用的。在这些注释中,我们认为所有的向量都是列向量。标量函数的向量默认为行向量。二阶导数表示法描述了标量函数g(y)= g(y1 · · · yn)的Hessian函数,g(y)= [g y1,· · ·,g yn ]
First order adjoints give the (first) derivatives of the cost functional with respect to the state. Second order adjoints give second derivatives of the cost functional with respect to the state. These derivatives are useful to speed up the optimization process in data assimilation, and to compute Hessian singular vectors. 1 Preliminaries In these notes we consider all vectors to be column vectors. Gradients of scalar functions are by default row vectors. Second derivative notation describes the Hessian of the scalar function, g(y) = g (y1 · · · yn) ⇒ ∂g ∂y = [ ∂g ∂y1 , · · · , ∂g ∂yn ]