Adaptive and Network-Aware Source Coding

源编码（数据压缩）的目标是给出数据的紧凑表示（例如，文本、图像、音频、视频以及诸如图形数据的其它非传统数据类型），根据这些数据可以完美地或者以感知上可接受的方式重构原始数据。随着大数据在我们社会中的扩散和大数据革命的到来，信源编码已经成为当代信息技术中的关键组成部分，因为它促进了大量数据的有效存储和远程访问，以及承载信息的信号在诸如互联网或无线网络的通信信道上的有效传输。拟议的研究将侧重于设计和分析新的源编码方法，以满足当代环境所带来的挑战，其中多个用户，通过不同容量的多个链接连接，希望通信或存储可能不是固定的和/或结构良好的数据。新的和实质性的研究挑战出现在方案的分析和设计，适应不断变化的信源统计，并纳入实际的限制内存，复杂性和延迟，对比更成熟的传统信源编码理论，这些约束通常被忽视。 本论文的研究目标分为两个方面：（1）系统研究单用户、分布式和双向网络中信源的最优零时延或有限时延码的性能和精细结构;（2）研究数据压缩和在线学习/分布式优化之间的丰富联系，以促进低复杂度的构建和分析，用于分布式设置中高度非平稳源的有损编码的自适应顺序方案，例如用于传感器网络的编码。 拟议研究的培训部分将提供2名硕士和3名博士。学生每年与刺激的研究挑战，并沉浸在信息和通信技术的重要课题。该研究有望提供有损信源编码的基本原理以及它的实际方面，如在通信网络中的多媒体信号的传输更深入的理解。

The goal of source coding (data compression) is to give a compact representation of data (e.g., text, image, audio, video, and other, non-traditional data types such as graphical data) from which the original data can be reconstructed either perfectly or in a perceptually acceptable manner. With the proliferation of large data sets in our society and the coming of the big data revolution, source coding has become a key component in contemporary information technologies as it facilitates the efficient storage and remote access of large volumes of data, as well as the efficient transmission of information-bearing signals over communication channels such as the Internet or wireless networks. The proposed research will focus on designing and analyzing novel source coding methods that meet the challenges posed by contemporary settings in which several users, connected via multiple links of different capacities, want to communicate or store data that may not be stationary and/or well structured. New and substantial research challenges arise in the analysis and design of schemes that adapt to changing source statistics and incorporate practical constraints on memory, complexity, and delay, contrasting the more established traditional theory of source coding where such constraints are usually disregarded. The research objectives are divided into two main themes: (1) The systematic study of the performance and finer structure of optimal zero- or limited-delay codes for information sources in the single user and distributed settings as well as for two-way networks; (2) The investigation of the rich connections between data compression and online learning/distributed optimization to facilitate the construction and analysis of low-complexity, adaptive sequential schemes for lossy coding of highly non-stationary sources in distributed settings, such as coding for sensor networks. The training component of the proposed research will provide 2 M.Sc. and 3 Ph.D. students each year with stimulating research challenges and immerse them in important current topics in information and communication technologies. The research is expected to provide a deeper understanding of the fundamental principles of lossy source coding as well as its practical aspects, such as the transmission of multimedia signals in communication networks.