U.S.-Mexico Cooperative Research: Long-Term Dynamics of Neural Network Approximations: Theory, Computation and Applications

小行星9503481 该美洲计划奖将支持普林斯顿大学Yannis Kevrekidis博士和墨西哥塞拉亚技术研究所Ramiro Rico-Martinez博士之间的合作研究项目，研究基于神经网络的模型的长期动态特性。 他们将研究所构建模型的长期预测的定性特性，并提出探索另一个内在模型特征，不可逆性，以澄清其在模型预测能力丧失中的作用及其在自适应识别和控制应用中的影响。 此外，他们计划开发，实施和测试新的方法，并研究相空间重建程序的属性。 将特别强调开发计算机工具，用于并行执行和分析模型。 拟议的研究将受益于Kevrekidis博士在非线性动力学方面的综合技能，以及Rico- Martinez博士在神经网络方面的综合技能。 ***

9503481 Kevrekidis This Americas Program award will support a collaborative research project between Dr. Yannis Kevrekidis, Princeton University, and Dr. Ramiro Rico-Martinez, Instituto Tecnologico de Celaya, Mexico, on the long-term dynamical properties of neural network-based models . They will study the qualitative properties of the long-term predictions of the constructed models and propose to explore another intrinsic model feature , noninvertibility, to clarify its role in the loss of model predictive capability and its impact in adaptive identification and control applications. In addition, they plan to develop, implement, and test new approaches , and study the properties of phase-space reconstruction procedures. Special emphasis will be placed on the development of computer tools for the parallel implementation and analysis of the models. The proposed research will benefit from the combined skills of Dr. Kevrekidis on non-linear dynamics, and Dr. Rico- Martinez in neural networks. ***