论文标题

部分可观测时空混沌系统的无模型预测

External univalence for second-order generalized algebraic theories

论文作者

Bocquet, Rafaël

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。

Voevodsky's univalence axiom is often motivated as a realization of the equivalence principle; the idea that equivalent mathematical structures satisfy the same properties. Indeed, in Homotopy Type Theory, properties and structures can be transported over type equivalences. However, we may wish to explain the equivalence principle without relying on the univalence axiom. For example, all type formers preserve equivalences in most type theories; thus it should be possible to transport structures over type equivalences even in non-univalent type theories. We define external univalence, a property of type theories (and more general second-order generalized algebraic theories) that captures the preservation of equivalences (or other homotopy relations). This property is defined syntactically, as the existence of identity types on the (syntactically defined) coclassifying (Sigma,Pi_rep)-CwF (also called generic model or walking model) of the theory. Semantically, it corresponds to the existence of some left semi-model structure on the category of models of the theory. We give syntactic conditions that can be used to check that a theory satisfies external univalence. We prove external univalence for some theories, such as the first-order generalized algebraic theory of categories, and dependent type theory with any standard choice of type formers and axioms, including identity types, Sigma-types, Pi-types, universes à la Tarski, the univalence axiom, the Uniqueness of Identity Proofs axiom, etc.

扫码加入交流群

加入微信交流群

微信交流群二维码

扫码加入学术交流群,获取更多资源