CIF: Small: New inroads into finite-length distributed information coding

该项目的目标是开发一种新的编码技术，用于形成许多网络构建块的三个基本多用户通信问题：(a)分布式源编码，(b)多重描述编码和(c)广播信道。这个项目是基于最近的两项初步调查。前者在分布式源代码编码中发现了一种新的编码现象。在后者中，为所有三个问题开发了一个新的统一代数编码框架，加强了它们之间的联系，从而表明新现象也在其他两个问题中发挥作用。众所周知，对于点对点通信问题（包括源编码和信道编码），随着码块长度的增加，性能或多或少地不断提高，直到无穷大。对于分布式源编码问题，采用一种特殊的量化和分码技术，在多个终端使用相同的码，结果表明，随着块长度的增加，性能先提高，然后达到平稳期，然后下降。因此，当代码的块长度为有限值而不是无穷大时，可以实现最佳性能。该项目通过开发基于这些代码的新框架来构建高效的编码和解码系统，解决了在一般网络上可靠传输和处理信息的问题。分析了这些代码在分布式源编码问题上的性能，并给出了性能极限的可计算边界。本项目使用该框架来解决广播信道问题和多重描述问题，使用三个问题之间存在的基本连接。该项目还追求具有计算效率的编码和解码算法的代码构建，以实现上述编码系统的实际实现。该项目将对以下方面产生影响：(1)不同研究领域之间的思想和概念交流；(2)下一代通信网络的设计和开发，如自组织无线和传感器网络；(3)通过将研究进展纳入课程开发计划，并努力招募代表性不足的少数民族和女性研究生，培养未来的工程师。

The objective of this project is to develop a new coding technique for three basic multiuser communication problems that form the building blocks of many networks: (a) distributed source coding, (b) multiple description coding and (c) broadcast channel. This project is based on two recent preliminary investigations. In the former, a new coding phenomenon was uncovered in distributed source coding. In the latter, a new unified algebraic coding framework for all three problems was developed that strengthened the connection among them, and thus suggesting that the new phenomenon is also at play in the other two. It is well-known that for point-to-point communication problem (both source coding and channel coding), as the block-length of the code is increased the performance keeps improving more or less all the way to infinity. For the problem of distributed source coding, with a particular class of technique called quantization and binning using identical codes at multiple terminals, it turns out that as the block-length is increased, the performance improves initially, reaches a plateau and then decreases after that. So the best performance is achieved when the block-length of the code is at some finite value but not infinity. This project addresses the problem of reliable transmission and processing of information over general networks by developing a new framework based on these codes to construct efficient encoding and decoding systems. The performance of these codes for the distributed source coding problem is analyzed and computable bounds on the performance limits are provided. This project uses this framework for the broadcast channel problem and the multiple description problem using a fundamental connection that exists among the three problems. This project also pursues code construction with computationally efficient encoding and decoding algorithms toward practical implementation of the coding systems developed above.This project will have an impact on (i) the exchange of ideas and concepts among different research fields, (ii) the design and development of the next generation of communication networks such as ad-hoc wireless and sensor networks, and (iii) education of future engineers by incorporating research advances into a curriculum development program and an effort toward recruitment of underrepresented minority and women graduate students.