N2GON: Neural Networks for Graph-of-Net with Position Awareness

Graphs, fundamental in modeling various research subjects such as computing networks, consist of nodes linked by edges. However, they typically function as components within larger structures in real-world scenarios, such as in protein-protein interactions where each protein is a graph in a larger network. This study delves into the Graph-of-Net (GON), a structure that extends the concept of traditional graphs by representing each node as a graph itself. It provides a multi-level perspective on the relationships between objects, encapsulating both the detailed structure of individual nodes and the broader network of dependencies. To learn node representations within the GON, we propose a position-aware neural network for Graph-of-Net which processes both intra-graph and inter-graph connections and incorporates additional data like node labels. Our model employs dual encoders and graph constructors to build and refine a constraint network, where nodes are adaptively arranged based on their positions, as determined by the network's constraint system. Our model demonstrates significant improvements over baselines in empirical evaluations on various datasets.

Many real-world systems, like biological interactions or even the internet, are complex networks. But what if each "node" or point in these networks is itself another intricate network? (For example, a protein in a Protein–protein interaction network is a complex graph of atoms). To address this, our research introduces a new way to see these systems, called "Graph-of-Net" (GoN). In a GoN, each node is recognized as its own detailed graph. We then developed an AI model, N2GON, specifically designed to learn from these multi-level structures. N2GON cleverly analyzes both the internal details of each "node-graph" and how these node-graphs are linked together in the overarching "net." It even considers their relative "positions" and distinct features to better understand their relationships.This approach allows for a much richer and more accurate understanding of complex, layered systems. For instance, it can help us better analyze how networks of proteins interact or how scientific papers, seen as text graphs, form citation networks. Our experiments show this method significantly improves our ability to analyze and interpret such intricate data compared to existing techniques.