SPP-CNN: An Efficient Framework for Network Robustness Prediction

Abstract:

SPP-CNN is an efficient method for network robustness approximation based on Spatial Pyramid Pooling, Its contributions are as follows:

• SPP-CNN is proposed, which has a wider tolerance to different input-data sizes than the CNN- and GNN-based approaches, while maintaining fast approximation speed like the CNN-based approaches. Together with the sorting strategy, the robustness prediction performance of SPP-CNN significantly outperforms the other CNN- and GNN-based approaches.
• SPP-CNN shows stronger generalizability than the other approaches on predicting the network robustness for datasets with unseen topologies and sizes.
• SPP-CNN demonstrates better performances than the other approaches on predicting the robustness of real-world networks, with consistent advantages for both synthetic and real-world networks.

Citation:

@article{Wu2023TCASI,
  author={Wu, Chengpei and Lou, Yang and Wang, Lin and Li, Junli and Li, Xiang and Chen,   Guanrong},
  journal={IEEE Transactions on Circuits and Systems I: Regular Papers}, 
  title={SPP-CNN: An Efficient Framework for Network Robustness Prediction}, 
  year={2023},
  volume={},
  number={},
  pages={1-13},
  doi={10.1109/TCSI.2023.3296602}
}

Download:

• PDF
• Code