Detection Models and Computation for Understanding the Frangibility of Complex Networks
1Deakin University, Melbourne, Australia
2China University of Mining and Technology, Xuzhou, China
3The University of Queensland, Brisbane, Australia
Detection Models and Computation for Understanding the Frangibility of Complex Networks
Description
A network is one of the best data models to represent human behavior and interactions. For instance, social network users can share their interests and news on social network platforms, traffic users can express their common interests and daily activities by trajectory networks, and telephone network users show their social interactions at periodic levels. In the area of network science, lots of research has been developed to investigate information flow and influence diffusion models in such complex networks. However, the frangibility of the network has not yet been fully explored. The frangibility of a network may be affected by multiple factors in the network, e.g., the removal of users may result in the disconnection of the network, or the high communication cost of the network; the addition of users may enrich the network so that the communication cost can be largely reduced; the changed paths of consuming network content may also vary the communication of the information flow in the network.
It is highly challenging to diagnose the frangibility of complex networks. Traditional works have paid great efforts to discover the influential users or seed users at different situations using metrics such as centrality, betweenness, and ego network, or using the linear threshold model, independent cascade model, and the weighted user-defined path models. However, these works still focused on the influence diffusion, rather than the frangibility of the networks in the dynamic environment. In addition, it is desirable for researchers to investigate the non-human defined metrics to uncover the frangibility of networks, with which the correlated key players can be retrieved. To solve this challenging issue, in machine learning, concept learning has become one of the primary research tools in small sample learning research. The concept learning strategy aims to perform recognition or form new concepts from a few observations though fast processing. Concept learning employs matching rules to associate the concepts in the concept system with small sample input. It is very helpful to perform cognition or complete a recognition task in data analytics. With the capability of small samples and the learned knowledge, it can help to advance the realistic key player discovery models and efficiently find the key players for different target criteria.
This Special Issue invites researchers working in the field of knowledge-based systems, data science, and artificial intelligence to submit original papers discussing and promoting ideas and practices about advanced complex network management and analytics technologies for the frangibility-driven key user discovery. Original research and review articles are welcome.
Potential topics include but are not limited to the following:
- Key user detection model with dynamic network change
- Concept learning from small samples to user profiling in networks
- Concept learning from small samples to attribute filling in networks
- Concept learning for event type disambiguation in networks
- New knowledge-driven concept learning across multiple networks
- Root cause diagnoses-based concept learning for network flow change
- Novel feature detection model for identifying the frangibility of networks
- Efficient structural hole computation in complex networks
- Top-k influential user detection in attributed networks
- Anchor vertex exploration to enrich the networks
- Dynamic network metric evaluation
- Community-level information diffusion with concept learning