605 ニューラルネットを用いた原子間相互作用データベース構築に関する基礎的研究 [in Japanese] Fundamental Research on Interatomic Potential Database Construction using Neural Network [in Japanese]
Access this Article
In this paper we apply a neural network to evaluation problem of interatomic potential that is determined by the arrangement of surrounding atoms. We examine the character of several proposed neural network structures, using simple atomic models. We focus mainly on the quasi-newton method and the backpropagation algorithm. Simulation results show that the quasi-newton method represents a gain concerning number of iteration of weights.
- The Proceedings of Conference of Kansai Branch
The Proceedings of Conference of Kansai Branch 2003.78(0), _6-9_-_6-10_, 2003
The Japan Society of Mechanical Engineers