【摘要】 深度学习技术已经广泛应用到大数据处理中,并在很多方面获得了可观的成绩.其中,自编码神经网络作为一种特征降维算法已被广大专家学者所应用.本文主要讨论一种改进的自动编码器——栈式降噪自编码神经网络(The Stacked Denoising Auto Encoder,SDAE),该算法使学习到的特征更加具有鲁棒性.并研究了该算法基于Re LU激活函数的中文短文本分类.与KNN,SVM,BP对比,无论召回率还是准确率,SDAE均优于KNN、BP、SVM.更多还原

【Abstract】 Deep learning technology has been widely used in big data processing and has achieved considerable achievements in many ways.One of these,the Auto Encoder has been used as a feature dimension reduction algorithm by many experts and scholars. We mainly discuss an improved algorithm—the Stacked Denoising Auto Encoder. This algorithm makes the features more robust its application in Chinese short text classification is discussed based on Re LU activation function.Compared with KNN,SVM and BP,SDAE is better on recall and accuracy rate.更多还原