Table 1.
Graph lifelong learning method comparison.
| Methods | Approach | |||
|---|---|---|---|---|
| Architectural | Rehearsal | Regularization | Reference | |
| Feature Graph Networks | Yes | No | No | Sarlin et al. (2020) and Zhou et al. (2022) |
| Hierarchical Prototype Networks | Yes | No | No | Li et al. (2023) and Zhang et al. (2023a) |
| Experience Replay GNN Frame work | No | Yes | No | Ahrabian et al. (2021a) and Zhou and Cao (2021b) |
| Lifelong Open-world Node Classification | No | Yes | No | Galke et al. (2021) and Zhang et al. (2022) |
| Disentangle-based Continual Graph Representation Learning | No | No | Yes | Kou et al. (2020) and Zhang et al. (2023b) |
| Graph Pseudo Incremental Learning | No | No | Yes | Tan et al. (2022) and Su et al. (2023) |
| Topology-aware Weight Preserving | No | No | Yes | Natali et al. (2020) and Liu et al. (2021) |
| Translation-based Knowledge Graph Embedding | No | No | Yes | Yoon et al. (2016) and Li et al. (2023) |
| Continual GNN | No | Yes | Yes | Han et al. (2020) and Wang et al. (2020) |
| Lifelong Dynamic Attributed Network Embedding | Yes | Yes | Yes | Li et al., 2017, Yoon et al. (2017), and Liu et al. (2021) |