CAREER: Human-Centric Big Network Embedding

网络嵌入是学习一种低维表示，以促进网络分析应用，包括节点分类和网络可视化。本项目旨在探索人类知识如何增强网络嵌入以及人类如何更好地理解结果的新方向。人类知识代表任何上下文信息或先验知识，这些信息或先验知识可能与该上下文中的学习嵌入相关。这项多学科研究的成功成果将推动领域专家利用人类知识交互、轻松地分析大网络数据，从而对各种信息系统的整体价值产生积极影响。综合数据科学教育计划旨在培养学生掌握关键但高度不可用的数据分析技术，吸引代表性不足的群体成员从事工程职业，并留住这些群体的成员。本项目的研究目标是开发一个以人为中心的框架，用于网络嵌入建模和整合人类知识，应对大网络带来的数据挑战，并实现网络嵌入结果的解释和交互。本项目开发了一系列不同于数据驱动方法的网络嵌入模型和算法，从各个方面分析网络数据。研究了多视图学习和深度结构化框架，将节点级、边缘级和社区级三种人类知识整合到一个统一的框架中。鉴于现实世界的网络可能包含异质、大规模和动态的人类知识，因此开发了相应的解决方案来处理这些问题。为了便于人类理解研究成果，本项目开发了全局和局部解释算法来解释网络嵌入和交互式学习算法来整合用户反馈。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。

Network embedding is to learn a low-dimensional representation to facilitate network analytics applications including node classification and network visualization. This project is to investigate a novel direction to explore how human knowledge could enhance network embedding and how the results could be better understood by human beings. Human knowledge represents any context information or prior knowledge that could be correlated to the learned embedding in this context. The successful outcome of this multidisciplinary research will lead to advances in enabling domain experts to interactively and easily analyze big network data with human knowledge, and thus positively impacting the overall value of various information systems. The integrated data science education program is to train students with crucial but highly unavailable data analytics technologies, to attract members of underrepresented groups to careers in engineering, and to retain members of those groups. The research goal of this project is to develop a human-centric framework for modeling and incorporating human knowledge in network embedding, tackling data challenges brought by big networks, as well as enabling interpretation and interaction of network embedding results. This project develops a series of network embedding models and algorithms, different from data-driven approaches, to analyze network data from various aspects. Multi-view learning and deep structured frameworks are investigated to integrate three types of human knowledge from the node-, edge- and community-level into a unified framework. Given the fact that real-world networks could contain heterogeneous, large-scale and dynamic human knowledge, corresponding solutions are developed to handle the problems. To facilitate human understanding of the research results, this project develops global and local interpretation algorithms to explain network embedding and interactive learning algorithms to integrate user feedback.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.