一种改进的字典学习方法在医疗疾病分析中的应用

针对传统字典学习算法忽略样本与字典原子之间关联信息及有限的训练集容易出现过拟合的现象,文中提出了通过加权机制及二范数约束的方法,建立样本与字典原子之间的权重关系,并对编码系数用二范数进行约束。采用留一交叉验证法在抑郁症数据集上比较了算法的精确度、灵敏度和错误率等指标。实验结果表明,改进的算法在医疗疾病分类中有良好的效果,分类精确度平均提高了21.87%,灵敏度和错误率等方面也有良好的表现。

An Improved Dictionary Learning Method for Medical Disease Analysis

This paper proposed the weighted mechanism to combine the sample with the dictionary atom based on the traditional dictionary learning algorithm, which ignored the relationship between the sample and the dictionary atom. At the same time, the l₂ norm regularization constraint was adopted to avoid over fitting on coding coefficients. The paper used the leave one out cross validation to compare the accuracy, sensitivity and mean error rate of the algorithm in the depression data sets. The results showed that the improved method had a good effect on the medical disease classification. The average classification accuracy was improved by 21.87%, and the sensitivity and mean error rate also displayed good performance.