| 摘要: |
| 提出了一种基于域理论的自适应谐振神经网络模型FTART2(field theory based adaptive resonance theory 2).该模型结合了自适应谐振理论和域理论的优点,学习速度快,归纳能力强,效率高,可以根据输入样本自适应地调整拓扑结构,克服了前馈型网络需要人为设置隐层神经元的缺点.基准测试表明,FTART2在学习精度和速度上都远远优于标准BP算法. |
| 关键词: 神经网络,机器学习,竞争学习,分类,自适应谐振理论,域理论. |
| DOI： |
| 分类号: |
| 基金项目:文本研究得到国家自然科学基金(No.69875006)和江苏省自然科学基金(No.BK9036)资助. |
|
| Field Theory Based Adaptive Resonance Neural Network Classifier |
|
ZHOU Zhi-hua,CHEN Zhao-qian,CHEN Shi-fu
|
| Abstract: |
| A field theory based adaptive resonance neural network model, FTART2, is proposed in this paper. FTART2 combines the advantages of the adaptive resonance theory and the field theory, and achieves fast learning, strong generality and high efficiency. Moreover, FTART2 can adaptively adjust its network topology so that the disadvantage of manually configuring hidden neurons of traditional feed-forward networks is avoided. Benchmark tests show that FTART2 achieves higher accuracy and faster speed than standard BP. |
| Key words: Neural network, machine learning, competitive learning, classification, adaptive resonance theory, field theory. |
