CRII: CIF: Models, Theories and Algorithms for Timeliness Optimization in Information-update Systems

在过去的二十年里，网络资源分配的理论基础和控制机制的发展取得了重大进展。这些新开发的理论和机制在吞吐量和延迟方面显著提高了网络性能。然而，对于需要实时信息更新的网络系统来说，优化吞吐量和延迟是不够的。需要在很大程度上扩展最先进的理论基础，以便将信息的及时性整合到网络控制机制的设计中。时效性优化的研究还处于起步阶段。预计该项目产生的新的理论成果和实际解决方案不仅将对信息理论和网络社区产生重大影响，而且将对数据库和机器学习社区产生重大影响。该项目将侧重于为本科生和K-12学生提供研究经验，招募和咨询代表性不足的学生，并参与课程开发活动。这项研究的目标是开发新的模型、理论和算法来优化信息更新系统的及时性性能。最近提出的一种被称为信息年龄或简称“年龄”的指标，将被用作研究及时性表现的关键指标。首先，本研究调查了信道编码对在有损信道上传输的信息的及时性的影响。其次，在一个新的环境下，研究了有界失效约束下的年龄最小化问题，其中信息可以被划分为多个具有部分更新的不相交单元。最后，本研究引入了一种新的拉取模型，其中目的地向源发送查询以拉取感兴趣的信息，并提出使用复制机制来优化时效性性能。

The last two decades have witnessed significant advances in the development of theoretical foundations and control mechanisms for network resource allocation. These newly developed theories and mechanisms have substantially improved network performance in terms of throughput and delay. However, optimizing throughput and delay is insufficient for networked systems that require real-time information update. The state-of-the-art theoretical foundations need to be largely expanded to integrate timeliness of information into the design of network control mechanisms. The research on timeliness optimization is still at its nascent stage. New theoretical results and practical solutions coming out of this project are expected to have a significant impact not only on information theory and networking community, but also on databases and machine learning community. This project will focus on providing research experiences to undergraduate and K-12 students, recruiting and advising underrepresented students, and engaging in curriculum development activities. The goal of this research is to develop new models, theories, and algorithms for optimizing timeliness performance in information-update systems. A recently proposed metric called age-of-information or simply "age", will be employed as a key metric to study timeliness performance. First, this research investigates the impact of channel coding on timeliness of information transmitted over a lossy channel. Second, this research studies the problem of age minimization under a bounded staleness constraint in a new setting where information can be partitioned into multiple disjoint units with partial updates. Finally, this research introduces a new Pull model where the destination sends queries to the sources to pull information of interest and proposes using replication schemes to optimize timeliness performance.