【摘要】 针对模糊神经网络分类器设计过程中所遇到的样本采样与标注过程耗时、代价大的问题,提出了一个新颖的模糊神经网络分类器主动学习方法,以最小-最大边界法以及确定样本的不确定性阈值两个新概念为主动样本选择准则,确保选择其中信息量尽可能大的样本进行标注,使得网络设计过程中对未标注样本的标注工作量和时间大为减少.实验结果表明,该方法与模糊神经网络的被动学习模型相比,训练样本数目大为减少,训练时间大大缩短.更多还原

【Abstract】 A novel approach to active learning based on fuzzy neural network classifier was proposed to solve the problems of time-consuming and costly sample collection and annotation. Two new concepts: min-max margin based approach and uncertainty threshold on samples were introduced as a rule of active sample selecting to guarantee that the most informative samples are annotated, thus reducing the annotation and time cost greatly. The experimental results show that the proposed method has better performance than passivity learning.更多还原