CCSS: Hyper-Graph Signal Processing for Multimedia Data Analysis in Cyber System Applications

数据科学（DS）和人工智能（AI）领域的研发活动浪潮正在推动一些最令人兴奋和改变游戏规则的技术进步。人工智能和物联网的广泛应用不断产生大量复杂的数据。这些新数据集比传统应用程序更大、更复杂。为了探索各种数据集的底层结构和相互作用，重要的几何信号处理工具（如图和超图模型）可以揭示新的见解，并在传感器网络、物联网、网络物理系统和多媒体数据分析中取得了广泛的成功。本研究项目以超图信号处理（hypergraph signal processing， HSP）的综合研究为中心。HSP能够捕捉信号或数据之间的多边关系和特征，在更广泛的实际应用中具有继承的分析能力。该项目代表了系统和全面的努力，以扩大几何信号处理的理论基础，并开发创新的解决方案，以响应大量涌入的现实生活数据集。一个目标是为数据分析建立一个定义良好且全面的HSP框架，与传统的图形信号处理兼容。该团队严格追求HSP的分析框架，并在特定应用中展示其优越的性能。PI和团队进一步提出了一个综合框架，系统地利用关键实际场景中的几何数据特征。研究团队计划开发创新的解决方案，用于分析和建模来自各种新应用的大型复杂数据集。解决关键数据分析主题，从这项工作中产生的创新可以直接影响网络系统中的数据压缩和分析。受HSP启发的大型数据集建模可以改变物联网和传感器网络的学习成果。这一研究成果将为信号处理和建模提供重要的理论基础。探索和发现新的数据结构和特征的能力将有助于更好地理解和优化各种数据科学和人工智能应用中的决策。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Waves of research and development activities on Data Science (DS) and Artificial Intelligence (AI) that are fueling some of the most exciting and game-changing technological advances. The widespread applications of AI and IoT generate huge volume of complex data incessantly. These new datasets are much bigger and more complex than in traditional applications. To explore the underlying structure and interaction of various datasets, important geometric signal processing tools such as graph and hypergraph models can reveal new insights and have found broad successes in sensor networks, IoT, cyber-physical systems, and multimedia data analysis. This research project centers on a comprehensive investigation of hypergraph signal processing (HSP). Capable of capturing multilateral relationships and features among signals or data, HSP provides inherit analytic strength in a wider range of practical applications. This project represents a systematic and comprehensive effort to broaden the theoretical foundation of geometric signal processing and to develop innovative solutions in response to the huge influx of real-life datasets. One goal is to establish a well-defined and comprehensive HSP framework for data analysis, compatible with traditional graph signal processing. The team rigorously pursues analytical frameworks for HSP and demonstrate its superior performance in specific applications. The PI and the team further present an integrative framework to systematically exploit geometric data features in key practical scenarios. The research team plans to develop innovative solutions for analyzing and modeling large and complex datasets from a variety of new applications. Addressing key data analytic topics, innovations arising from this work can directly impact data compression and analysis in cyber-systems. HSP inspired modeling of large data collections can transform learning outcomes in IoT and sensor networks. This research outcomes shall contribute substantially to the theoretical foundation of signal processing and modeling. The ability to explore and discover new data structures and features will enable better understanding and optimized decision making in various data science and artificial intelligence application.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.