Geometric Graph Neural Networks

Thesis project repository https://arxiv.org/pdf/2211.03232.pdf

Pending Meeting Tasks (TO DO)

• Añadir reset de parámetros
• Hacer experimento de entrenamiento con y sin batches normalizando por el nro de FLOPS
• Revisar qué hace la función de plotting
• Investigar sobre edge prediction
• Investigar sobre qué hace la función de pérdida de pytorch
• Realizar experimento para medir sensibilidad de poolings sobre grafos construidos a mano
• Implementar k-disGNN [2] y geoGNN [1]

Work Log

---

• 12-jan

  • 👥 Meeting

• 11-jan

  • 📚 Read [] for better undestanding of edge prediction

• 9-jan:

  • 👥 Meeting
  • 📚 Code review of [2]
  • ✨ Added graph_prediciont_regresion.ipynb
  • 📚 Read PyTorch Geometric documentation for transform and filter PyG datasets

• 8-jan:

  • ✨ Added node_classification.ipynb, graph_prediciont_binary.ipynb and basicGNN.py
  • 🔥 Remove old code
  • ⚡ Permanently install Python libraries on Paperspace

• 7-jan:

  • 📚 Read PyTorch Geometric documentation on datasets, layars, models and run notebooks examples on Paperspace using graphcore IPU
  • 🔥 Deprecate data folder

• 4-jan:

  • 👥 Meeting

• 3-jan:

  • 📚 Read [9, 10, 11, 12, 13] for graph pooling and readout

• 2-jan:

  • 📚 Read [5, 6, 7, 8] for graph pooling and readout
  • ✨ Adapt graphSAGE for PROTEINS dataset
  • ♻️ Refactor of GAT. Work in progress

• 1-jan:

  • ♻️ Review and refactor of graphSAGE [4]. Add MessagePassing class

• 28-dic:

  • Review Mojo🔥 programming languange documentation for possible High efficient GNN implementation.

• 26-dic:

  • ✨ Added QM9.ipynb
  • ♻️ Refactor of datasets scripts by integrate PyTorch Geometric lib. for dataset managing.
  • ✨ Added MD17.ipynb, this .ipynb describes the MD17 benzene dataset
  • ✨ Added DD.ipnb, this .ipynb describes the DD dataset
  • ✨ Connect Github repo with Paperspace workspace

• 25-dic:

  • ✨ Added Cora.ipynb, this .ipynb describes the cora dataset

---

Bibliography

1. Rose, V. D., Kozachinskiy, A., Rojas, C., Petrache, M., & Barceló, P. (2023). Three iterations of $(d-1)$-WL test distinguish non isometric clouds of $d$-dimensional points. arXiv

2. Li, Z., Wang, X., Huang, Y., & Zhang, M. (2023). Is Distance Matrix Enough for Geometric Deep Learning?. arXiv

3. Morris, C., Ritzert, M., Fey, M., Hamilton, W. L., Lenssen, J. E., Rattan, G., & Grohe, M. (2018). Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks. arXiv

4. Hamilton, W. L., Ying, R., & Leskovec, J. (2017). Inductive Representation Learning on Large Graphs. arXiv

5. Liu, C., Zhan, Y., Wu, J., Li, C., Du, B., Hu, W., Liu, T., & Tao, D. (2022). Graph Pooling for Graph Neural Networks: Progress, Challenges, and Opportunities. arXiv

6. Grattarola, D., Zambon, D., Bianchi, F. M., & Alippi, C. (2021). Understanding Pooling in Graph Neural Networks. arXiv

7. Ying, R., You, J., Morris, C., Ren, X., Hamilton, W. L., & Leskovec, J. (2018). Hierarchical Graph Representation Learning with Differentiable Pooling arXiv

8. Buterez, D., Janet, J. P., Kiddle, S. J., Oglic, D., & Liò, P. (2022). Graph Neural Networks with Adaptive Readouts. arXiv

9. Ju, W., Fang, Z., Gu, Y., Liu, Z., Long, Q., Qiao, Z., Qin, Y., Shen, J., Sun, F., Xiao, Z., Yang, J., Yuan, J., Zhao, Y., Luo, X., & Zhang, M. (2023). A Comprehensive Survey on Deep Graph Representation Learning (Version 2). arXiv

10. Hamilton, William L. (2023). Graph Representation Learning. Synthesis Lectures on Artificial Intelligence and Machine Learning. link

11. Mesquita, D., Souza, A. H., & Kaski, S. (2020). Rethinking pooling in graph neural networks (Version 1). arXiv

12. Pal, S., Malekmohammadi, S., Regol, F., Zhang, Y., Xu, Y., & Coates, M. (2020). Non-Parametric Graph Learning for Bayesian Graph Neural Networks. arXiv

13. Zhang, Z., Bu, J., Ester, M., Zhang, J., Yao, C., Yu, Z., & Wang, C. (2019). Hierarchical Graph Pooling with Structure Learning. arXiv