Reduction and Classification of Input Parameters for Large-scale Simulations

概要

論文の詳細を見る
An integrated management system for simulation cycles has been developed and performed. However, the existing management system often requires complex operation for data management. We investigate an autonomous agent system that advises the simulation users about management operation for large-scale simulations. Specifically, we develop an input parameter customizing agent which is a part of the agent system. As the first prototype of the agent, we analyze input parameter sets for large-scale simulations at JAERI Kansai Research Establishment Advanced Photon Research Center. In this paper, we propose an input parameter analysis method; input parameter sets are reduced to smaller dimensions using Principal Component Analysis (PCA), and classified using Learning Vector Quantization (LVQ) according to simulation types. The validation of the input parameter analysis method is presented by experiments.
一般社団法人情報処理学会の論文
2003-06-24