Learning adaptive neighborhoods for graph neural networks
Graph convolutional networks (GCNs) enable end-to-end learning on graph structured data. However, many works assume a given graph structure. When the input graph is noisy or unavailable, one approach is to construct or learn a latent graph structure. These methods typically fix the choice of node de...
| Main Authors: | , , , |
|---|---|
| 格式: | Conference item |
| 語言: | English |
| 出版: |
IEEE
2024
|
| _version_ | 1826316050653773824 |
|---|---|
| author | Saha, A Mendez, O Russell, C Bowden, R |
| author_facet | Saha, A Mendez, O Russell, C Bowden, R |
| author_sort | Saha, A |
| collection | OXFORD |
| description | Graph convolutional networks (GCNs) enable end-to-end learning on graph structured data. However, many works assume a given graph structure. When the input graph is noisy or unavailable, one approach is to construct or learn a latent graph structure. These methods typically fix the choice of node degree for the entire graph, which is suboptimal. Instead, we propose a novel end-to-end differentiable graph generator which builds graph topologies where each node selects both its neighborhood and its size. Our module can be readily integrated into existing pipelines involving graph convolution operations, replacing the predetermined or existing adjacency matrix with one that is learned, and optimized, as part of the general objective. As such it is applicable to any GCN. We integrate our module into trajectory prediction, point cloud classification and node classification pipelines resulting in improved accuracy over other structure-learning methods across a wide range of datasets and GCN backbones. We will release the code. |
| first_indexed | 2024-03-07T08:01:42Z |
| format | Conference item |
| id | oxford-uuid:301c56db-82b1-43db-8c14-cb2b8b444499 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-12-09T03:39:05Z |
| publishDate | 2024 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:301c56db-82b1-43db-8c14-cb2b8b4444992024-12-06T10:14:09ZLearning adaptive neighborhoods for graph neural networksConference itemhttp://purl.org/coar/resource_type/c_5794uuid:301c56db-82b1-43db-8c14-cb2b8b444499EnglishSymplectic ElementsIEEE2024Saha, AMendez, ORussell, CBowden, RGraph convolutional networks (GCNs) enable end-to-end learning on graph structured data. However, many works assume a given graph structure. When the input graph is noisy or unavailable, one approach is to construct or learn a latent graph structure. These methods typically fix the choice of node degree for the entire graph, which is suboptimal. Instead, we propose a novel end-to-end differentiable graph generator which builds graph topologies where each node selects both its neighborhood and its size. Our module can be readily integrated into existing pipelines involving graph convolution operations, replacing the predetermined or existing adjacency matrix with one that is learned, and optimized, as part of the general objective. As such it is applicable to any GCN. We integrate our module into trajectory prediction, point cloud classification and node classification pipelines resulting in improved accuracy over other structure-learning methods across a wide range of datasets and GCN backbones. We will release the code. |
| spellingShingle | Saha, A Mendez, O Russell, C Bowden, R Learning adaptive neighborhoods for graph neural networks |
| title | Learning adaptive neighborhoods for graph neural networks |
| title_full | Learning adaptive neighborhoods for graph neural networks |
| title_fullStr | Learning adaptive neighborhoods for graph neural networks |
| title_full_unstemmed | Learning adaptive neighborhoods for graph neural networks |
| title_short | Learning adaptive neighborhoods for graph neural networks |
| title_sort | learning adaptive neighborhoods for graph neural networks |
| work_keys_str_mv | AT sahaa learningadaptiveneighborhoodsforgraphneuralnetworks AT mendezo learningadaptiveneighborhoodsforgraphneuralnetworks AT russellc learningadaptiveneighborhoodsforgraphneuralnetworks AT bowdenr learningadaptiveneighborhoodsforgraphneuralnetworks |