Towards Deeper Graph Neural Networks
問題： 一層的GCN layer只考慮直接鄰域，加深模型層數可獲得更大的感受野receptive fields，但性能卻會變壞。 以前認爲是over-smoothing問題導致的。 本文 增加層性能變壞是因爲GCN模型裏兩個關鍵部分——representation transformation和propagation——的糾纏導致的。 分離這兩個部分可以解決該問題。 模型： 進行特徵變換， 堆疊層
相關文章
- 1. 圖神經網絡論文閱讀(二十) Towards Deeper Graph Neural Networks
- 2. Applications of Graph Neural Networks
- 3. Inceptionism: Going Deeper into Neural Networks[譯]
- 4. GMNN: Graph Markov Neural Networks
- 5. CS224W-08-Graph Neural Networks
- 6. Adaptive Graph Convolutional Neural Networks
- 7. 筆記《Graph Neural Tangent Kernel: Fusing Graph Neural Networks with Graph Kernels》-NeurIPS 2019
- 8. Few-shot Learning with Graph Neural Networks
- 9. Few-Shot Learning with Graph Neural Networks
- 10. Overlapping Community Detection with Graph Neural Networks
- 更多相關文章...
- • Git 標籤 - Git 教程
- • Git 查看提交歷史 - Git 教程
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. 圖神經網絡論文閱讀(二十) Towards Deeper Graph Neural Networks
- 2. Applications of Graph Neural Networks
- 3. Inceptionism: Going Deeper into Neural Networks[譯]
- 4. GMNN: Graph Markov Neural Networks
- 5. CS224W-08-Graph Neural Networks
- 6. Adaptive Graph Convolutional Neural Networks
- 7. 筆記《Graph Neural Tangent Kernel: Fusing Graph Neural Networks with Graph Kernels》-NeurIPS 2019
- 8. Few-shot Learning with Graph Neural Networks
- 9. Few-Shot Learning with Graph Neural Networks
- 10. Overlapping Community Detection with Graph Neural Networks