(2020.10) 分子のグラフ表現と機械学習: Graph Neural Networks (GNNs) とは？
0 likes1,983 views
The document discusses various aspects of molecular graphs, including chemical structures and properties such as dipole moments, energy levels, and atomization energies. It references multiple sources related to graph neural networks and their applications in deep learning, particularly in contexts like retrosynthesis prediction and molecular graph generation. Additionally, it includes links to tutorials, datasets, and relevant academic papers in the field.
![node[0] node[1]
Edge 0
Edge 4
0 1](https://image.slidesharecdn.com/csjfesta2020-201022221704/85/2020-10-Graph-Neural-Networks-GNNs-13-320.jpg)
(2020.10) 分子のグラフ表現と機械学習: Graph Neural Networks (GNNs) とは？
- 4. ...
- 5. J’aime la musique I love music
- 10. Sci Data 1, 140022 (2014). https://doi.org/10.1038/sdata.2014.22
- 12. Dipole moment Isotropic polarizability Highest occupied molecular orbital (HOMO) energy Lowest unoccupied molecular orbital (LUMO) energy Gap between HOMO and LUMO Electronic spatial extent Zero point vibrational energy Internal energy at 0K Internal energy at 298.15K Enthalpy at 298.15K Free energy at 298.15K Heat capavity at 298.15K Atomization energy at 0K Atomization energy at 298.15K Atomization enthalpy at 298.15K Atomization free energy at 298.15K Rotational constant A Rotational constant B Rotational constant C N O C C C C H H H H H
- 13. node[0] node[1] Edge 0 Edge 4 0 1
- 16. H C N O F (one-hot encoding) 0 1 2 3 4 5 6 7 8 9 10 https://pytorch-geometric.readthedocs.io/en/latest/_modules/torch_geometric/datasets/qm9.html N O C C C C H H H H H
- 34. 5 0.20.3 0.41 -2
- 47. 5 0.20.3 0.41 -2 0.17 0.05 1.0 0.95 8.7 1.4 1 -0.99 0.3 0.17 5 -1 1
- 48. 5 0.20.3 0.41 -2 0.17 0.05 1.0 0.95 8.7 1.4 1 -0.99 0.3 0.17 5 -1 1 5.9
- 52. 5.9 5.9 5.9 1 5 -1 0.17 -1 0.3 -0.99 1 29.5 -1.00 1.75 1 -0.99 0.3 0.17 5 -1 1 5.9 = = = × × × × × × × ×
- 53. 5 0.20.3 0.41 -2 0.17 0.05 1.0 0.95 8.7 1.4 1 -0.99 0.3 0.17 5 -1 1 5.9
- 57. x = x1 x2 x1 x2 1 1 w0 ji w00 i wkj 1 y import torch.nn as nn nn.Sequential(nn.Linear(2, 3), nn.ReLU(), nn.Linear(3, 2), nn.ReLu(), nn.Linear(2, 1)) 1
- 58. N O C C C C H H H H H N O C C C C H H H H H MLP (sum, mean or max) MLP
- 59. 1 2 3 4 5 6 0 1 0 0 0 0 1 0 1 1 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 1 0 1 2 3 4 5 6 1 2 3 4 5 61 2 3 4 5 6 0 1 0 0 0 0 1 0 1 1 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 0 1 1 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 0 1 1 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 1 0 = = = 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 3 4 5 2 + + +
- 64. N O C C C C H H H H H (sum, mean or max)
- 72. https://ai.tencent.com/ailab/ml/KDD-Deep-Graph-Learning.html • KDD2020 Tutorial Deep Graph Learning: Foundations, Advances and Applications http://cse.msu.edu/~mayao4/dlg_book/index.html • Book "Deep Learning on Graphs" • CS224W: Machine Learning with Graphs (Stanford) http://web.stanford.edu/class/cs224w/ https://youtu.be/-UjytpbqX4A
- 77. Generalization and Representational Limits of Graph Neural Networks Garg, Jegelka, Jaakkola (ICML2020) PairNorm: Tackling Oversmoothing in GNNs Zhao, Akoglu (ICLR2020) Towards Deeper Graph Neural Networks with Differentiable Group Normalization Zhou, Huang, Li, Zha, Chen, Hu (NeurIPS2020) Strategies for Pre-training Graph Neural Networks Hu, Liu, Gomes, Zitnik, Liang, Pande, Leskovec (ICLR2020) GROVER: Self-Supervised Message Passing Transformer on Large-scale Molecular Graphs Rong, Bian, Xu, Xie, Wei, Huang, Huang (NeurIPS2020) DeepGCNs: Can GCNs Go as Deep as CNNs? Li, Müller, Thabet, Ghanem (ICCV2019)
- 78. Path Integral Based Convolution and Pooling for Graph Neural Networks Ma, Xuan, Guang Wang, Li, Liò (NeurIPS2020) Hierarchical Generation of Molecular Graphs using Structural Motifs Jin, Barzilay, Jaakkola (ICML2020) GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation Shi, Xu, Zhu, Zhang, Zhang, Tang (ICLR2020) RetroXpert: Decompose Retrosynthesis Prediction like A Chemist Yan, Ding, Zhao, Zheng, Yang, Yu, Huang (NeurIPS2020) A Graph to Graphs Framework for Retrosynthesis Prediction Shi, Xu, Guo, Zhang, Tang (ICML2020) Rethinking Pooling in Graph Neural Networks Mesquita, Souza, Kaski (NeurIPS2020)