CIF: Small: A Stochastic Approximation Approach to Network Communications with Feedback

具有反馈和中继的跨网络通信是信息通信理论中一个重要的、具有挑战性的、很大程度上开放的问题。系统和可扩展的通信算法将提高可靠性，降低编码复杂性或延迟，通过利用多样性赋予通信方案鲁棒性，并在某些情况下提高可实现的通信速率。由于所有现代通信系统都采用多种反馈机制，因此更好地理解通信反馈是必要的。本研究的重点是通过共识算法和更一般的随机逼近算法来对这一主题进行更深入的研究。这种方法为与现代通信系统极端相关的情况（例如，具有噪声反馈的网络场景）产生可扩展和健壮的算法。利用了信息论和分布式动态系统的问题、技术和工具的融合，对这两个领域都有潜在的变革性影响。除了分析技术和模拟之外，该团队还利用设备和经验在基于软件定义无线电的无线网络测试台上实现新的通信算法。对具有噪声反馈的网络间通信问题的任何影响都将立即应用于大多数现代通信系统。此外，本研究进一步将这两个方面和研究团体统一起来，形成一个更普遍的信息理论——既考虑数据的传输（经典信息理论），也考虑数据的利用（经典动力系统）。强调在会议中组织特别会议，提供新的研究生课程，包括本科生的实验研究，指导少数民族学生，并进一步向对工程感兴趣的高中生推广，是该项目的一个组成部分。

Communication across networks with feedback and relays is an important, challenging, and largely open problem in information and communication theory. Systematic and scalable communication algorithms will enhance reliability, decrease coding complexity or delay, impart robustness to communication schemes by exploiting diversity, and in some cases increase the achievable communication rates. Since all modern communication systems employ multiple feedback mechanisms, a better understanding of communication feedback is imperative.This research focuses on a deeper study of this topic by viewing it through the lenses of consensus algorithms and more general stochastic approximation algorithms. This approach yields scalable and robust algorithms for cases of extreme relevance to modern communication systems, e.g., network scenarios with noisy feedback. A confluence of problems, techniques and tools from information theory and distributed dynamic systems are utilized, with a potential transformative impact on both fields. In addition to analytical techniques and simulations, the team also utilizes facilities and experience realizing novel communication algorithms in a wireless network testbed based upon software-defined radios.Any impact on the problem of communication across networks with noisy feedback will have immediate applications to most modern communication systems. Moreover, this research furthers the unification of the two aspects and research communities relevant to a more general information theory - one that considers both the transmission of data (classical information theory), as well as its utilization (classical dynamical systems). An emphasis on organizing special sessions in conferences, offering new graduate courses, including undergraduates in the experimental research, mentoring minority students, and furthering outreach to high school students interested in engineering is an integral part of the project.