在普通非负矩阵分解(NMF)方法基础上提出了3个二进制约束非负矩阵分解(3bNMF)算法,对分解矩阵和恢复矩阵元素增加了二进制数的约束,从而更适合对二进制数据进行处理.分别给出了3bNMF算法在数字矩阵分解和有噪声情况下汉字偏旁部首提取中的应用实例,并与普通NMF方法所处理的结果进行了比较.
Based on Lee and Seung's NMF, 3bNMF method is proposed by considering binary constraints for factorized base matrix, presentation matrix and restoration matrix. 3bNMF is applied to digital matrix factorization and base structure extraction respectively from Chinese characters with noise. The results are compared with NMF.