Tractable frameworks for complex network modelling

许多现代应用程序生成具有网络结构的大规模数据，该网络结构通常包括时态组件、不同的连接模式以及边、节点和子图级别的信息。找到易于处理的统计框架来利用这些数据可能会产生重要的社会效益，例如，在网络安全应用（例如入侵检测，“假新闻”），医学（例如生物网络，遗传学），“人工智能”应用（例如推荐系统，情感分析，自然语言处理）等。该博士将寻求开发此类解决方案，最初专注于嵌入方法，即首先将复杂的离散数据结构转换为点云，然后通过更标准的统计和机器学习技术（如聚类）进行后续分析。

Many modern applications generate large-scale data with a network structure which often includes a temporal component, different modes of connection, and information at edge, node and subgraph levels. Finding tractable statistical frameworks for exploiting such data could have important societal benefits, for example, in cyber-security applications (e.g. intrusion detection, 'fake news'), medicine (e.g. biological networks, genetics), 'artificial intelligence' applications (e.g. recommender systems, sentiment analysis, natural language processing) and more. This PhD will seek to develop such solutions, initially focussing on the approach of embedding, whereby a complex discrete data structure is first transformed into a point cloud, allowing subsequent analysis by more standard statistical and machine-learning techniques such as clustering.