论文部分内容阅读
文中提出了一种具有抗噪音能力的增量式混合学习算法ＩＨＭＣＡＰ．该算法将基于概率论的符号学习与神经网络学习相结合，通过引入ＦＴＡＲＴ神经网络，不仅实现了两种不同思维层次的靠近，还成功地解决了符号学习与神经网络学习精度之间的均衡性问题．其独特的增量学习机制不仅使得它只需进行一遍增量学习即可完成对新增示例的学习，还使该算法具有较好的抗噪音能力，从而可以应用于实时在线学习任务．
In this paper, an anti-noise ability incremental hybrid learning algorithm IHMCAP is proposed. The algorithm combines symbolic learning based on probability theory with neural network learning. By introducing FTART neural network, not only the approach of two different thinking levels is realized, but also the balance between the learning accuracy of symbolic learning and neural network is successfully solved problem. Its unique incremental learning mechanism not only makes it possible to perform incremental learning with just one increment of learning, but also has better anti-noise ability so that it can be applied to real-time online learning tasks.