抄録
A hybrid fuzzy-neuro classifier that extracts rules in terms of polyhedrons in the input space is proposed. The network uses a fuzzy disjunctive normal form in its hidden layer to effectively map polyhedral regions, which are gradually adjusted during learning, to category labels. The major advantage of the present method lies in that it is quite simple in architecture, every layer enjoys a clear fuzzy logical interpretation, and the number of rules needed is very few. The results of classification experiments seem to be quite promising.
本文言語 | English |
---|---|
ページ | 577-580 |
ページ数 | 4 |
出版ステータス | Published - 2000 1月 1 |
イベント | FUZZ-IEEE 2000: 9th IEEE International Conference on Fuzzy Systems - San Antonio, TX, USA 継続期間: 2000 5月 7 → 2000 5月 10 |
Other
Other | FUZZ-IEEE 2000: 9th IEEE International Conference on Fuzzy Systems |
---|---|
City | San Antonio, TX, USA |
Period | 00/5/7 → 00/5/10 |
ASJC Scopus subject areas
- ソフトウェア
- 理論的コンピュータサイエンス
- 人工知能
- 応用数学