Integrating multiple information sources and evaluating joint impact in network inference

Develop methods to effectively incorporate diverse node attributes and multiple interaction types within attributed multilayer networks and rigorously evaluate their collective impact on downstream network inference tasks.

Background

Existing probabilistic models for attributed networks typically handle only single-layer or homogeneous data types (e.g., one interaction type and one categorical attribute), making it difficult to represent real-world scenarios with heterogeneous information. The authors highlight that combining various sources of information and assessing their joint effect on inference tasks (such as community detection and prediction) remains unresolved in the literature, motivating their development of a more flexible approach.