PAGCL: An unsupervised graph poisoned attack for graph contrastive learning model.
Qing LiZiyue WangZehao LiPublished in: Future Gener. Comput. Syst. (2023)
Keyphrases
- graph representation
- graph model
- prior knowledge
- bipartite graph
- mathematical model
- structural learning
- fully connected
- learning models
- learning process
- dependency graph
- reinforcement learning
- em algorithm
- graph structure
- supervised learning
- directed acyclic graph
- graph theory
- probabilistic model
- graph matching
- connected components
- machine learning
- computational model
- online learning
- probability distribution
- tree structure
- active learning
- learning systems
- unsupervised learning
- random walk
- learning scheme
- objective function
- graph theoretic
- learning mechanism
- similarity measure