粗集理论在神经网络模式识别中的应用

Application of Rough Sets Theory in NN Pattern Recognition

基于粗糙集理论在神经网络模式识别中,利用粗糙集将采集的原始数据作为学习样本,量化样本属性值,组织决策表,将样本的条件和决策属性建为二维表并约简条件属性,仅保留影响分类的重要属性。将决策规则中必要条件属性神经网络化,通过粗糙集和神经网络并行学习实现最小决策规则,直到由最少属性构成的决策规则网络能正确划分所有测试集样本为止。

In NN pattern recognition based on rough set theory, gathered original data was used as learning stylebook by rough set, the attribute of stylebook was measured and decision-making table was built with the rules of pattern recognition. The 2D table was set up with condition and decision-making attributes of stylebook, and condition attribute was reduced and some of important attributes were saved. Necessary condition attribute in picking-up rules were trained in the BP neural network and rough set at the same time to get Least decision-making rules, until all the tested stylebook are rightly assigned to NN according to decision-making rules founded by least attributes.