深度学习在数据压缩中的应用研究

发布时间：2018-09-14 14:54

【摘要】：随着信息化时代的来临,数据如何被快速地传输和高效地存储变成一个引人关注的话题,而数据压缩技术恰好是解决这一问题的有效方法。数据压缩是指在保证信息的完整性或者数据的重要信息不丢失的同时,减少数据的体积,从而提高其传输、存储和处理效率的一种技术。数据压缩技术日益被人们关注的原因有两点:一是存储数据的设备的容量的是有一定限制的,为了能够在有限的空间存储更多的数据就需要对原始数据进行压缩。二是信息技术的高速发展对数据传递速度的要求越来越高。在数据传递过程中数据包越小传递的速度就越快,所以在为了获得更快的传递速度和更高的处理效率,在满足使用者对信息要求的情况下,数据量越小越好。随着大数据时代的来临,数据的存储和传输面临着越来越多的问题。由于数据的体积呈现爆发式增长、数据的结构越来越复杂,使得传统的数据压缩算法面临着严峻的挑战。恰是在此背景下,本文提出了把深度学习技术和聚类算法相结合的模型用于数据压缩的研究。从实验结果来看,混合模型在取得高压缩比的同时也保存了大量的重要信息,使得解压后的数据能够较好的表示出原始数据,非常适用于结构复杂数据的有损压缩。本文还在数据预处理阶段使用了数据补全和数据校正技术,保证了数据的完整性,减少了噪声数据对实验的影响。在把深度学习技术用于智能电表数据压缩的研究中,本文还使用了无损编码技术应用于有损压缩后数据的进一步压缩。最后本文采用了交叉验证技术验证了模型的性能,实验结果表明这一混合模型是可靠、稳定的。
[Abstract]:With the advent of the information age, how data can be transmitted quickly and stored efficiently has become a hot topic, and data compression technology is just an effective way to solve this problem. Data compression technology has been paid more and more attention to for two reasons: first, the capacity of data storage devices is limited, in order to be able to store more data in a limited space, it is necessary to compress the original data. second, the rapid development of information technology on data. In order to achieve faster transmission speed and higher processing efficiency, the smaller the amount of data, the better. With the advent of the era of large data, data storage and transmission are facing. Because of the explosive growth of data volume and the complexity of data structure, traditional data compression algorithms are facing severe challenges. Under this background, this paper proposes a model which combines depth learning technology with clustering algorithm for data compression. The model not only obtains high compression ratio but also saves a lot of important information, so that the decompressed data can better represent the original data, which is very suitable for lossy compression of complex structure data. In the research of applying depth learning technology to data compression of smart meters, lossless coding technology is used to further compress the data after lossy compression. Finally, the performance of the model is verified by cross-validation technology. The experimental results show that the hybrid model is reliable and stable.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP181

【参考文献】