基于栈式自编码融合极限学习机的药品鉴别  

Drug identification based on stacked auto encoders fusing extreme learning machine

作　　者：张卫东 路皓翔 甘博瑞 杨辉华 ZHANG Wei-dong;LU Hao-xiang;GAN Bo-rui;YANG Hui-hua(College of Computer and Information Security,Guilin University of Electronic Technology,Guilin 541004,China;College of Automation,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]桂林电子科技大学计算机与信息安全学院,广西桂林541004 [2]北京邮电大学自动化学院,北京100876

出　　处：《计算机工程与设计》2019年第2期545-549,561,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(21365008;61105004);广西壮族自治区自然科学基金项目(2013GXNSFBA019279)

摘　　要：近红外光谱数据维度较高,传统的特征提取方法不足以提取更高层次的抽象特征,为此提出一种栈式自编码融合极限学习机的药品鉴别方法,利用ELM代替SAE的反向微调和Softmax分类阶段,减少了SAE的训练时间,提高了SAE的应用能力。以不同厂商生产的非铝塑包装的头孢克肟片药品的近红外光谱为实例,在不同规模的数据集下,验证该算法,并与其它机器学习方法进行对比。实验结果表明,SAE-ELM减少了SAE的训练时间,具有较高分类准确率和稳定性。Near infrared spectroscopy data are high dimensional,and the traditional feature extraction method is not enough to extract higher-level abstract features.To solve the problems,a drug discrimination method based on stacked auto encoders(SAE-ELM)fusion extreme learning machine was proposed.By introducing ELM instead of the back fine-tuning and Softmax classification stages of SAE,the training time was reduced and the ability of applications of SAE was improved.Using NIRS of cefixime tablet with non-Aluminum packaged by different manufacturers with different sizes of data sets as an example to verify the proposed method,and compared to others machine methods,the result shows that SAE-ELM not only reduces the training time,but shows high classification accuracy and stability.

关 键 词：抽象特征 近红外光谱 药品鉴别 栈式自编码 极限学习机 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]