论文标题
部分可观测时空混沌系统的无模型预测
Global Cellular Automata GCA -- A Massively Parallel Computing Model
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。
The Global Cellular Automata (GCA) Model is a generalization of the Cellular Automata (CA) Model. The GCA model consists of a collection of cells which change their states depending on the states of their neighbors, like in the classical CA model. In generalization of the CA model, the neighbors are no longer fixed and local, they are variable and global. In the basic GCA model, a cell is structured into a data part and a pointer part. The pointer part consists of several pointers that hold addresses to global neighbors. The data rule defines the new data state, and the pointer rule define the new pointer states. The cell's state is synchronously or asynchronously updated using the new data and new pointer states. Thereby the global neighbors can be changed from generation to generation. Similar to the CA model, only the own cell's state is modified. Thereby write conflicts cannot occur, all cells can work in parallel which makes it a massively parallel model. The GCA model is related to the CROW (concurrent read owners write) model, a specific PRAM (parallel random access machine) model. Therefore many of the well-studied PRAM algorithms can be transformed into GCA algorithms. Moreover, the GCA model allows to describe a large number of data parallel applications in a suitable way. The GCA model can easily be implemented in software, efficiently interpreted on standard parallel architectures, and synthesized / configured into special hardware target architectures. This article reviews the model, applications, and hardware architectures.
