Extending the Conventional Information Theory: Theory and Designs of Layered Partially Decodable Codes

摘要

This project is aimed at extending the conventional information theory, which is based on the idea of reliably transmitting all data bits through communication channels, by differentiating different types of data and providing different levels of error protections and thus allowing different parts of the data to be decoded at different nodes in the network. The main challenge addressed in this project is to develop new notions of partial information, and new coding techniques to encode different types of data together, efficiently. Over a 3-year period of research, significant progresses have been made in the following areas: (1) new analysis tools have been developed to establish the fundamental performance limits of layered codes, in terms of the tradeoff between the rates and reliability levels of different parts of data encoded together; (2) new applications of embedded coding in networks, including efficient transmission of network control messages and channel state information, have been investigated, showing significant improvements of the overall performance; (3) new connections to information geometry is observed, which leads to novel approaches to classical network information theory problems, a few instances have been reported.