摘要
A fast learning method for morphological neural networks which are formulated based on max plus algebra, is proposed. The proposed fast learning method is deduced from the idempotent properties of max plus algebra and cost matrices. Through experiments using artificial training data sets, it is conformed that the computation time of the proposed method is decreased into 60.9-68.6% and 40.9 - 48.3% of that of the conventional one, under the condition that the number of middle layer nodes is 4 and 8, respectively.
| 源语言 | 英语 |
|---|---|
| 出版状态 | 已出版 - 2005 |
| 已对外发布 | 是 |
| 活动 | 3rd Serbian-Hungarian Joint Symposium on Intelligent Systems, SISY 2005 - Subotica, 塞尔维亚 期限: 31 8月 2005 → 1 9月 2005 |
会议
| 会议 | 3rd Serbian-Hungarian Joint Symposium on Intelligent Systems, SISY 2005 |
|---|---|
| 国家/地区 | 塞尔维亚 |
| 市 | Subotica |
| 时期 | 31/08/05 → 1/09/05 |
指纹
探究 'A fast learning method using cost matrices for morphological neural networks' 的科研主题。它们共同构成独一无二的指纹。引用此
Nobuhara, H., Bede, B., & Hirota, K. (2005). A fast learning method using cost matrices for morphological neural networks. 论文发表于 3rd Serbian-Hungarian Joint Symposium on Intelligent Systems, SISY 2005, Subotica, 塞尔维亚.