论文部分内容阅读
深度学习是当前智能识别、数据挖掘等领域最重要的研究方向,通过组合低层特征,形成更加抽象的高层表示属性类别或特征,以发现数据的分布式特征表示.数据降维是深度学习过程中最为常见的一种过程,通过降维,能够去除数据间的相关性,便于提取更为有用的数据特征,提升识别率,加快识别速度.数据降维过程中,必然导致数据信息的损失,如何统计运用这个信息损失,目前还少有相关文献进行研究.通过对栈式自编码器深度学习算法进行研究,提出一种深度学习降维信息损失度量方法,将香农信息理论运用到降维信息损失度量中,计算深度学习降维过程中信息损失量,并研究其与算法性能的关系,为深度学习算法的改进提供数据支撑.
Depth learning is the most important research direction in the field of intelligent recognition and data mining at present, and forms a more abstract high-level representation attribute category or feature by combining low-level features to find the distributed feature representation of data.Dimensions reduction is the process of deep learning The most common kind of process, through the dimension reduction, can remove the correlation between data, facilitate the extraction of more useful data features, improve the recognition rate and speed up the recognition speed.During the data dimension reduction, it will inevitably lead to the loss of data information, how to There are few relevant literatures to study the application of this information loss.Through the study of deep learning algorithm for stack self-encoder, a new measure method of information loss for deep learning dimensionality reduction is proposed, which applies Shannon information theory to the loss of information loss Metric, the amount of information loss during deep learning dimensionality reduction is calculated, and its relationship with algorithm performance is studied to provide data support for the improvement of deep learning algorithm.