| 初始权值优化技术在SOM网络中的应用 |
| |
| 引用本文： | 彭雅琴,陈俊,宫宁生.初始权值优化技术在SOM网络中的应用[J].计算机工程与设计,2008,29(23). |
| |
| 作者姓名： | 彭雅琴 陈俊 宫宁生 |
| |
| 作者单位： | 1. 南京工业大学,信息科学与工程学院,江苏,南京,210009
2. 南京工业大学,信息科学与工程学院,江苏,南京,210009;南京航空航天大学,信息科学与工程学院,江苏,南京210016 |
| |
| 摘 要： | SOM网络是一种无导师学习方法,被广泛应用于各个领域.网络的性能受很多因素影响,如样本的选择、网络结构、初始权值的选定等.针对网络初始权值选取的不确定性问题,提出了覆盖权值初始化方法来优化SOM网络的初始权值:该方法从样本入手,并通过"覆盖"方法得出初始权值,仿真实验结果证明了此方法能有效的提高网络的识别率和稳定性.
|
| 关 键 词： | 权值优化 SOM网络 样本分布 归一化方法 权值分布 |
Implementation of optical weights initialization technology in SOM network |
| |
| PENG Ya-qin,CHEN Jun,GONG Ning-sheng.Implementation of optical weights initialization technology in SOM network[J].Computer Engineering and Design,2008,29(23). |
| |
| Authors: | PENG Ya-qin CHEN Jun GONG Ning-sheng |
| |
| Affiliation: | PENG Ya-qin1,CHEN Jun1,GONG Ning-sheng1,2(1.College of Information Science , Engineering,Nanjing University of Technology,Nanjing 210009,China,2.College of Information Science , Engineering,Nanjing University of Aeronautic , Astronautics,Nanjing 210016,China) |
| |
| Abstract: | SOM network is one of the unsupervised learning methods, which is widely applied in various fields.The performance of the SOM network is affected by many factors such as the sample selection, network structure, initial weight and so on.In order to solve the uncertain problem of the initial weight selection, a covering initialization theory is proposed, which is employed to optimize the initial weight of the SOM network.The method starts with training samples and acquires the initial weights by the way of ov... |
| |
| Keywords: | weight optimization SOM network samples distributing normalization method weight distributing |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
