基于改进的多层降噪自编码算法临床分类诊断研究

被引:9
作者
胡帅 [1 ]
袁志勇 [1 ]
肖玲 [2 ]
王惠玲 [2 ]
王高华 [2 ]
机构
[1] 武汉大学计算机学院
[2] 武汉大学人民医院
关键词
深度学习; 多层降噪自编码; 元代价; 分类诊断; 代价敏感; 不均衡;
D O I
暂无
中图分类号
TP181 [自动推理、机器学习]; R44 [诊断学];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ; 100208 ;
摘要
针对临床分类诊断中普遍存在的样本不均衡、错分代价不同、大量无标签样本和测量误差等特点,引入了机器学习中较新的研究成果——多层降噪自编码(stacked denoising autoencoders,SDA)神经网络,并与欠采样局部更新的元代价(metacost)算法相结合,对SDA神经网络进行了改进,使组合模型具有代价敏感、降低不均衡性、有效利用无标签样本、抗噪声的特性。实验中将改进的SDA神经网络与SOFTMAX回归、反向传播(back propagation,BP)神经网络、支持向量机(support vector machine,SVM)、传统多层自编码(stacked autoencoders,SAE)神经网络,以及传统SDA神经网络等作了比较。实验结果表明,改进的SDA神经网络的准确率、ROC曲线下面积等均优于其他模型,提高了分类模型的辅助诊断性能。
引用
收藏
页码:1417 / 1420
页数:4
相关论文
共 18 条
[1]  
LIBSVM[J] . Chih-Chung Chang,Chih-Jen Lin. &nbspACM Transactions on Intelligent Systems and Technology (TIST) . 2011 (3)
[2]   Bagging predictors [J].
Breiman, L .
MACHINE LEARNING, 1996, 24 (02) :123-140
[3]   INDUCTIVE AND BAYESIAN LEARNING IN MEDICAL DIAGNOSIS [J].
KONONENKO, I .
APPLIED ARTIFICIAL INTELLIGENCE, 1993, 7 (04) :317-337
[4]  
MetaCost:a general method for making classifiers costsensitive. Domingos,P. Data Mining and Knowledge Discovery . 1999
[5]  
A fast learning algorithm for deep belief nets. Hinton Geoffrey E,Osindero Simon,Teh Yee-Whye. Neural Computation . 2006
[6]  
A Simple Generalisation of the Area Under the ROC Curve for Multiple Class Classification Problems. David J Hand,Robert J Till. Machine Learning . 2001
[7]  
Learning deep architectures for AI. Bengio,Yoshua. Foundations and Trends in Machine Learning . 2009
[8]  
On the expressive power of deep architectures. BENGIO Y,DELALLEAU 0. Algorithmic Learning Theory . 2011
[9]  
AUC: A Statistically Consistent and More Discriminating Measure than Accuracy, Proc. C.X. Ling,J. Huang,and H. Zhang. IJCAI’03 . 2003
[10]  
Prediction as a candidate for learning deep hierarchical models of data. Palm R B. Journal of Women s Health . 2012