论文部分内容阅读
IHMCAP(incrementalhybridmulti-conceptsacquisitionprocedure)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(fieldtheory-basedadaptiveresonancetheory)神经网络,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,实现了两种不同思维层次的靠近.该算法采用一种独特的增量学习机制,当增加新的实例时,只需进行一遍增量学习,调整原结构,不必重新生成判定树和神经网络,即可提高学习精度,速度快,效率高.同时,这种增量学习机制还可以降低算法对噪音数据的敏感度,从而使IHMCAP可以应用于实时在线学习任务.
IHMCAP (incremental hybrid multi-concepts acquisitionprocedure) algorithm combines symbolic learning based on probability theory with neural network learning, successfully introduces the FTTH neural network, and successfully solves the problem of the balance between the symbolic learning and neural network learning precision, To achieve the two different levels of thinking close. The algorithm uses a unique increment learning mechanism. When a new instance is added, incremental learning and adjustment of the original structure can be performed without having to regenerate decision trees and neural networks to improve learning accuracy, speed, and efficiency high. At the same time, this incremental learning mechanism also reduces the sensitivity of the algorithm to noise data, enabling IHMCAP to be applied to real-time online learning tasks.