ITR/AP: A COMPUTATIONAL INFRASTRUCTURE FOR RELIABLE COMPUTER SIMULATIONS

EIA-0205181 J.Tinsley Oden of Texas大学，奥斯汀ITR/AP：用于可靠计算机模拟的编译基础设施本项目的研究将开发用于验证的数学和计算过程，类似于用于验证的过程，并通过实际和计算实验对这些过程进行评估。拟议的核查和确认综合框架的核心是分层建模的概念。分层建模是对基本模型的系统粗化，基本模型是先验已知的包含所有信息的基本模型，这些信息可以从给定的实验中提取到可量化的范围内。目标是通过在基本模型中引入近似来确定包含实验信息内容的计算最简单的模型。该过程需要评估由基本模型的每个近似引入的误差的能力、捕获模型和实验两者的语义内容的能力、比较实验和模拟的信息内容的能力、以及实现实现模型序列的一系列计算系统的能力，从而可以实现和评估概念过程。

EIA-0205181 J. Tinsley Oden University of Texas, Austin ITR/AP: A COMPILATION INFRASTRUCTURE FOR RELIABLE COMPUTER SIMULATIONSThe research of this project will develop mathematical and computational processes for validation analogous to those used for verification and evaluate these processes by actual and computational experiments. At the core of the proposed integrated framework for verification and validation lies the concept of hierarchical modeling. Hierarchical modeling is a systematic coarsening of mathematical models from a base model which is known a priori to contain all of the information, which can be extracted from a given experiment to within quantifiable bounds. The goal is to determine, by introducing approximations into the base model, the computationally simplest model which contains the information content of the experiment. This process requires an ability to evaluate the error introduced by each approximation to the base model, a capability for capturing the semantic content of both models and experiments, a capability for comparing the information content of the experiments and simulations and a capability for implementing a family of computational systems implementing the sequence of models so that the conceptual process can be realized and evaluated.