Boost then Convolve: Gradient Boosting Meets Graph Neural Networks
arXiv:2101.08543
Abstract
Graph neural networks (GNNs) are powerful models that have been successful in various graph representation learning tasks. Whereas gradient boosted decision trees (GBDT) often outperform other machine learning methods when faced with heterogeneous tabular data. But what approach should be used for graphs with tabular node features? Previous GNN models have mostly focused on networks with homogeneous sparse features and, as we show, are suboptimal in the heterogeneous setting. In this work, we propose a novel architecture that trains GBDT and GNN jointly to get the best of both worlds: the GBDT model deals with heterogeneous features, while GNN accounts for the graph structure. Our model benefits from end-to-end optimization by allowing new trees to fit the gradient updates of GNN. With an extensive experimental comparison to the leading GBDT and GNN models, we demonstrate a significant increase in performance on a variety of graphs with tabular features. The code is available: https://github.com/nd7141/bgnn.
ICLR 2021: camera-ready
References in corpus (12)
- Fast Graph Representation Learning with PyTorch Geometric
- VERSE: Versatile Graph Embeddings from Similarity Measures
- Attention-based Graph Neural Network for Semi-supervised Learning
- Strategies for Pre-training Graph Neural Networks
- Deep Neural Decision Trees
- Multi-Layered Gradient Boosting Decision Trees
- Residual Correlation in Graph Neural Network Regression
- Heterogeneous Deep Graph Infomax
- Deep Graph Matching Consensus
- Gradient Boosting Neural Networks: GrowNet
- Spatially-Aware Graph Neural Networks for Relational Behavior Forecasting from Sensor Data
- AdaGCN: Adaboosting Graph Convolutional Networks into Deep Models