论文部分内容阅读
CC4神经网络是一种三层前馈网络的新型角分类(corner classification)训练算法,原用于元搜索引擎Anvish的文档分类.当各文档之间的规模接近时,CC4神经网络有较好的分类效果.然而当文档之间规模差别较大时,其分类性能较差.针对这一问题,本文意图扩展原始CC4神经网络,达到对文档有效分类的效果.为此,提出了一种基于MDS-NN的数据索引方法,将每一文档映射至k维空间数据点,并尽可能多地保持原始文档之间的距离信息.其次,通过将索引信息变换为CC4神经网络接受的0,1序列,实现对CC4神经网络的扩展,使其能够接受索引信息作为输入.实验结果表明对相互之间规模差别较大的文档,扩展CC4神经网络的性能优于原始CC4神经网络的性能.同时,扩展CC4神经网络的分类精度与文档索引方法有密切关系.
The CC4 neural network is a new type of corner classification training algorithm for the three-layer feedforward network, which was originally used for the document classification of meta search engine Anvish.When the scale of each document is close, the CC4 neural network has good However, the classification performance is poor when there are large differences in the sizes of documents.To solve this problem, this paper intends to extend the original CC4 neural network to achieve the effective classification of documents.Therefore, a new MDS -NN data indexing method, each document is mapped to k-dimensional space data points, and as much as possible to maintain the distance between the original document information.Secondly, by transforming the index information into CC1N0 networks accepted by the 0,1 sequence , Which can expand the CC4 neural network and make it accept the index information as input.The experimental results show that the performance of the extended CC4 neural network is superior to that of the original CC4 neural network for the documents with large difference in scale between each other.At the same time, CC4 neural network classification accuracy and document indexing methods are closely related.