论文部分内容阅读
根据自适应谐振理论提出了半监督学习自适应谐振理论系统.在该系统中取消了一般半监督学习算法中假定已知数据概率分布的条件限制,利用自适应谐振理论的稳定性和可塑性,使其具有非常强的学习新模式和纠正错误能力.为了提高系统自适应性能力,将警戒参数设置为动态变化。实验结果表明半监督学习自适应谐振理论系统的性能优于判别式CEM算法,特别是在含有噪音和新模式数据情况下,其优势更为明显.
According to the theory of adaptive resonance, a semi-supervised learning adaptive resonance theory system is proposed, in which the conditional limitations of probability distribution of assumed data in general semi-supervised learning algorithm are eliminated. By using the stability and plasticity of adaptive resonance theory, It has a very strong learning new model and error correction capabilities.In order to improve the system adaptive ability, the alert parameters are set to dynamic changes. Experimental results show that the performance of semi-supervised learning adaptive resonance theory system is better than discriminative CEM algorithm, especially in the case of noise and new mode data.