论文标题
部分可观测时空混沌系统的无模型预测
Moments and tails of Lq-valued chaoses based on independent variables with log-concave tails
论文作者
论文摘要
储层计算是预测湍流的有力工具，其简单的架构具有处理大型系统的计算效率。然而，其实现通常需要完整的状态向量测量和系统非线性知识。我们使用非线性投影函数将系统测量扩展到高维空间，然后将其输入到储层中以获得预测。我们展示了这种储层计算网络在时空混沌系统上的应用，该系统模拟了湍流的若干特征。我们表明，使用径向基函数作为非线性投影器，即使只有部分观测并且不知道控制方程，也能稳健地捕捉复杂的系统非线性。最后，我们表明，当测量稀疏、不完整且带有噪声，甚至控制方程变得不准确时，我们的网络仍然可以产生相当准确的预测，从而为实际湍流系统的无模型预测铺平了道路。
We derive a lower bound for moments of random chaoses of order two with coefficients in arbitrary Banach space F generated by independent symmetric random variables with logarithmically concave tails (which is probably two-sided). We also provide two upper bounds for moments of such chaoses when F = L_q. The first is true under the additional subgaussanity assumption. The second one does not require additional assumptions but is not optimal in general. Both upper bounds are sufficient for obtaining two-sided moment estimates for chaoses with values in Lq generated by Weibull random variables with shape parameter greater or equal to 1.
